sunday

Published on November 2016 | Categories: Documents | Downloads: 162 | Comments: 0 | Views: 4191
of 72
Download PDF   Embed   Report

Comments

Content

TECHNICAL SESSIONS
How to Navigate the Technical Sessions
There are four primary resources to help you understand and navigate the Technical Sessions: • This Technical Session listing, which provides the most detailed information. The listing is presented chronologically by day/time, showing each session and the papers/abstracts/authors within each session. • The Session Chair, Author, and Session indices provide cross-reference assistance (pages 345-376). • The map and floor plans on pages 42 and 43 show you where technical session tracks are located. • The “Master Track Schedule” is on pages 44-47. This is an overview of the tracks (general topic areas) and when/where they are scheduled.

Sunday, 8:00am - 9:30am
I SA01
Geometric Optimization
Sponsor: Optimization/ Network and Combinatorial Optimization Sponsored Session
Chair: Kasturi Varadarajan, University of Iowa, 101E MacLean Hall, Iowa City, IA, United States, [email protected] 1 - Robust Shape Fitting via Peeling and Grating Coresets Pankaj Agarwal, Duke University, Department of Computer Science, Levine Science Research Center D315, Durham, NC, 27708, United States, [email protected]
Let P be a set of n points in d-dimensional Euclidean space. A (k,eps)-kernel is a subset of P that, for every direction, epsilon-approximates the directional width of P, when k “outliers’’ can be ignored in that direction. We prove that small (k,eps)-kernels exist and describe efficient algorithms for computing them. Such kernels are instrumental in solving shape-fitting problems with k outliers. This is joint work with Sariel Har-Peled and Hai Yu.

Quickest Way to Find Your Own Session
Use the Author Index (pages 349-367) — the session code for your presentation(s) will be shown along with the room location (abbreviated). You can also refer to the full session listing for the room location of your session(s).

2 - Randomized Algorithms for Geometric Set Cover Ken Clarkson, Distinguished Member of Technical Staff, Bell Laboratories, Lucent Technologies, 600 Mountain Avenue, Room 2C-526, Murray Hill, NJ, 07974, United States, [email protected]
While in general the set cover problem is hard to approximate, in the geometric setting it is possible, for a large class of problems, to obtain constant-factor approximations. I will describe an approach to do this, by applying random sampling to solve one particular easy version of the problem, that of finding “epsilon-nets”, and then to leverage that solution to greater generality. Applications include covering a region with disks and guarding simple terrains. Joint with K. Varadarajan.

The Session Codes

SB01
The day of the week

Track number. Coordinates with the room locations shown in the Master Track Schedule. Room locations are also indicated in the listing for each session. Time Block. Matches the time blocks shown in the Master Track Schedule.

3 - Efficient Subspace Approximation Algorithms Kasturi Varadarajan, University of Iowa, 101E MacLean Hall, Iowa City, IA, United States, [email protected], N D Shyamalkumar
We consider the problem of finding the optimal k-dimensional subspace that fits a set P of n points in d-dimensional Euclidean space. The quality of the fit of a subspace is measured by the sum of the distances of the n points to the subspace. We present approximation algorithms for this problem with running time linear in n and d and exponential in k and the approximation parameter. We also consider other related notions of fit.

Time Blocks
A— B— C— D— 8:00am - 9:30am 10:00am - 11:30am 1:30pm - 3:00pm 3:30pm - 5:00pm

4 - Thick Paths and Minimum-Cost Flows in Geometric Domains Valentin Polishchuk, PhD Candidate, Stony Brook University, Stony Brook, NY, 11794, United States, [email protected], Joseph Mitchell
Planning non-crossing routes amidst polygonal obstacles for multiple moving objects has been a central problem in robotics, air traffic management, VLSI routing, and other fields. We provide efficient algorithms for finding a set of shortest thick paths between pairs of terminals on the boundary of a polygonal domain. We apply our thick paths techniques to obtain the first algorithmic results for the minimum-cost continuous flow problem in geometric domains.

On Wednesday, D block is 3:30-5:30pm. Plenaries, keynotes and lunch breaks are interspersed among the technical session time blocks.

I SA02
Joint Session Open-Source/ICS: Applications of the Nonlinear Optimization Software IPOPT I
Cluster: Open-Source Software: Open Source, Open Standards, Open Data, INFORMS Computing Society Invited Session
Chair: Andreas Waechter, Research Staff Member, IBM TJ Watson Research Center, 1101 Kitchawan Road, Yorktown Heights, NY, 10598, United States, [email protected] 1 - Large-Scale Nonlinear Parameter Estimation with IPOPT Victor M. Zavala, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected], Lorenz Biegler
Previously intractable DAE-constrained estimation problems for complex process models have been considered. Although the related NLPs are highly nonlinear with possibly non-unique solutions, IPOPT has shown to be efficient in their solution. NLPs with up to 69,000 constraints and 530 degrees of freedom have been solved in a few minutes.

Room Locations/Tracks
All tracks and technical sessions are held in the Convention Center. Room numbers are shown on the Master Track Schedule and in the technical session listing. All 300 rooms and Ballroom B are on Level 3 of the Center; all 400 rooms are on Level 4.

Quick Reference
Don’t miss the “Quick Reference,” a separate flier you received in your registration packet. It includes the Master Track Schedule and floor plans, providing a quick, portable summary of the meeting

51

SA03

INFORMS PITTSBURGH — 2006 I SA04
Managing Supply Chain Uncertainties
Cluster: Supply Chain and Operations Engineering Invited Session
Chair: Yusen Xia, Assistant Professor, Georgia State University, Atlanta, GA, 30303, United States, [email protected] 1 - Supply Chain Revenue Management Yusen Xia, Assistant Professor, Georgia State University, Atlanta, GA, 30303, United States, [email protected], Qiying Hu
In this talk, we study revenue management for a one-supplier and one-retailer supply chain where consumers can buy either from the supplier or from the retailer. Both centralized and decentralized settings are investigated.

2 - Spacecraft Trajectory Optimization Juan Arrieta, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected], Lorenz Biegler
The optimal thrusting sequence that will transfer a spacecraft between two points in the Solar System can be obtained from the solution to a large, nonconvex optimal control problem. The NLP resulting from direct transcription is solved with IPOPT. We present realistic examples: an orbit transfer and an Earth-Moon trajectory.

3 - IPOPT in a Software Package for Dynamic Optimization Yi-Dong Lang, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected], Andreas Waechter, Lorenz Biegler
A software package, DynoPC, uses IPOPT as the NLP solver to solve discretized dynamic optimization problems simultaneously. It invokes ADOL-C to evaluate constraints, Jacobians and Hessians required by IPOPT, based on user-supplied models. The package has been updated according to new developments of IPOPT, and tested on challenging examples.

2 - A Procurement Model with Supply Uncertainty Haresh Gurnani, University of Miami, 417B Jenkins Building, Coral Gables, FL, 33124, United States, [email protected], Tridip Ray, Saibal Ray
With the rapid expansion of global B2B commerce, newer suppliers with cheaper but possibly unreliable technologies enter the market place to win orders from firms by beating the price of their perfectly reliable (but expensive) competitors. The dilemma facing purchasing firms is the allocation of the tender across suppliers of varying supply reliability. The cases of perfect, imperfect, and incomplete information will be discussed.

I SA03
Interior Point Methods for Linear Programming
Sponsor: Optimization/ Linear Programming and Complementarity Sponsored Session
Chair: Hande Benson, Assistant Professor, Drexel University, Department of Decision Sciences, 3141 Chestnut Street, Philadelphia, PA, 19104, United States, [email protected] 1 - Solving Linear Optimization Problems with MOSEK Bo Jensen, MOSEK ApS, C/O Symbion Science Park, Fruebjergvej 3, Copenhagen, 2100, Denmark, [email protected]
The software package MOSEK is capable of solving large-scale sparse linear optimization problems using either an interior-point, a primal simplex or a dual simplex algorithm. The aim of this talk is to present the optimizers and the recent advances in their implementation. Moreover, we will present numerical results demonstrating the optimizers performance.

3 - Performance Aspects of Collaborative Planning Forecasting and Replenishment (CPFR) Technologies Craig Hill, Assistant Professor, Georgia State University, Dept. of Managerial Sciences, 35 Broad Street, 10th Floor, Atlanta, GA, 30302, United States, [email protected], Yusen Xia, Peter Zhang
The CPFR initiative is a relatively new management paradigm that supports supply chain members making joint decisions on their mutual supply chain issues. These issues and decisions include such things as inventory management and product design. This research uses empirical data to focus on the performance ramifications to companies that use this program.

2 - Complexity Results for Path Following Algorithms for Linear Programming Renato Monteiro, Professor, ISyE, Georgia Institute of Technology, Atlanta, GA, United States, [email protected], Takashi Tsuchiya, Guanghui Lan
In this talk, we discuss complexities of path following algorithms for linear programming which take into account the geometry of the central path. We survey old iteration-complexity results for the the Vavasis-Ye and the MTY predictor-corrector algorithms. We also present complexity results for a new predictor-corrector trust region algorithm. We end the talk by surveying results about the geometric structure of the central path in terms of its curvature.

4 - Make or Buy: Impact of Information on Optimal Production/Outsourcing Policies with Limited Capacity Liwen Chen, PhD Student, The University of Texas at Austin, 1 University Station B6500, IROM Department, Austin, TX, 78712, United States, [email protected], Genaro Gutierrez
We study a production/outsourcing problem with limited capacity and outsourcing option using a formal Bayesian framework. We show find the optimal production/outsourcing policy. We provide monotonicity results of the critical points w.r.t. the production capacity and the information acquiring time. We find that the knowledge that we are going to get demand information later will change our operational decisions. We also explore mismatch cost, outsourcing option value.

3 - A Constraint Reduced IPM for Convex Quadratic Programming, with Application to SVM Training Jin Hyuk Jung, PhD Student, University of Maryland, Department of Computer Science, College Park, MD, 20742, United States, [email protected], Andre Tits, Dianne O’Leary
We propose a constraint reduction scheme for a convex QP with m inequality constraints and n variables,mdddn. We solve the normal equations with a reduced matrix, H + A’_Q diag(z/s)_Q A_Q, for a well chosen index set Q (adaptively shrunk) thus reducing the computational cost per iteration by a factor close to |Q|/m. We prove global, and local quadratic, convergence. We apply our scheme to support vector machine (SVM) training (primal QP formulation), with highly promising results.

I SA05
Computational Developments in Stochastic Integer Programming
Sponsor: Optimization/ Stochastic Programming Sponsored Session
Chair: Miguel Lejeune, Assistant Professor, Carnegie Mellon University, 5000 Forbes Avenue, Tepper School of Business, Pittsburgh, PA, 15213, United States, [email protected] 1 - Building Solutions for 2-Stage and Multi-Stage Stochastic Linear Programs Using Xpress-MP Nitin Verma, Dash Optimization, 560 Sylvan Avenue, Englewood Cliffs, NJ, United States, [email protected]
In this talk we demonstrate how one could use the modeling, programming, optimization, and visualization tools available in Xpress-MP to build solutions for stochastic problems. Specifically, we will show how to easily build comprehensive solutions — for both analyses and deployment purposes — in a relatively short frame of time using Xpress- SP together with Xpress-IVE, and XAD.

4 - Exact Primal-Dual Regularization of Linear Programs Dominique Orban, Ecole Polytechnique de Montreal, CP 6079 Succ. Centre-Ville, Montreal, QC, H3C-3A7, Canada, [email protected], Michael Friedlander
In the framework of linear programming, we propose a theoretical justification for regularizing the linear systems used to compute search directions when the latter are (nearly) rank deficient. We present a primal-dual infeasible algorithm for LPs with explicit primal and dual regularization. The regularization is termed exact to stress that, although the LP is regularized, we are still able to recover a solution of the original LP, independently of the values of the regularization parameters.

2 - An Exact Solution Approach for Stochastic Integer Portfolio Optimization Problems Miguel Lejeune, Assistant Professor, Carnegie Mellon University, 5000 Forbes Avenue, Tepper School of Business, Pittsburgh, PA, 15213, United States, [email protected], Pierre Bonami
We develop an exact solution method for portfolio optimization problems in which uncertainty in the estimate of the expected return and real-life market restriction modeled with integer constraints are simultaneously considered. We use a non-linear branch-and-bound algorithm and implement a new branching priority rule. Computational results on problems containing up to 200 assets are reported and benchmarked with others methods and solvers.

52

INFORMS PITTSBURGH— 2006
3 - Sharp Bound For Probabilities in a Distribution with Given Shape Information Ersoy Subasi, RUTCOR, Rutgers University, 640 Bartholomew Road, Piscataway, NJ, 08854, United States, [email protected], Mine Subasi, András Prékopa
The binomial moment problem with finite, preassigned supports and given shapes of the distribution is formulated and used to obtain sharp lower and upper bounds for probabilities that exactly r out of n events occurs as well as probabilities that at least r out of n events occurs. The bounds are based on the knowledge of some of the binomial moments of the number of events which occur. Numerical examples and applications in reliability and finance are presented.

SA08

4 - A Constraint-Programming Approach for Scheduling with Alternative Process Plans Andrew Davenport, IBM, IBM T.J.Watson Research Center, 1101 Kitchawan Road, PO Box 218, Yorktown Heights, NY, 10598, United States, [email protected]
Manufacturing scheduling problems often involve jobs which can be scheduled using one of a number of different process routes through the manufacturing facility, as well as preferences on which process routes are selected. We present techniques and results for scheduling with alternative process routes using constraint programming approaches.

4 - MIPing the Probability Set Covering Problem Vineet Goyal, PhD Student, Tepper School of Business, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected], Anureet Saxena, Miguel Lejeune
We consider a probabilistic set covering problem where the right hand side is a random binary vector and the covering constraint must be satisfied with some probability. The problem can be formulated as a disjunctive program over the set of p-efficient points, and is solved with a branch and bound algorithm. In this paper, we give a novel MIP formulation of the probability constraint using only few additional variables. This approach overcomes the exponential enumeration of p-efficient points.

I SA07
New Books in Business Statistics
Cluster: New Books Invited Session
Chair: Brett Saraniti, Hawaii Pacific University, 1060 Bishop Street, #402, Honolulu, HI, United States, [email protected] 1 - Presenting Business Statistics with “Concise Managerial Statistics” Kellie Keeling, Assistant Professor, Virginia Tech, Business Information Technology(0235), Pamplin 1007, Blacksburg, VA, 24061-0235, United States, [email protected]
Concise Managerial Statistics is an introduction to Business Statistics text integrating the use of computers with statistics, by describing the concepts and applications while giving users the opportunity to carry out computer-generated solutions using SPSS, Excel, and MINITAB. The text includes Excel Add-ins that expand Excel’s power in graphing and regression analysis. The cases, chapter examples, and exercises focus on the interpretation of statistical graphs and statistical conclusions.

5 - Maximization of a Strongly Unimodal Multivariate Discrete Distribution Mine Subasi, RUTCOR, Rutgers University, 640 Bartholomew Rd, Piscataway, NJ, 08854, United States, [email protected], Ersoy Subasi, András Prékopa
A dual type linear programming algorithm is presented to find the maximum of a strongly unimodal multivariate discrete distribution. Numerical examples are presented.

I SA06
Applications of Constraint Programming
Sponsor: INFORMS Computing Society/Constraint Programming Sponsored Session
Chair: Andrew Davenport, IBM, IBM T.J.Watson Research Center, 1101 Kitchawan Road, PO Box 218, Yorktown Heights, NY, 10598, United States, [email protected] 1 - Distributed Management of Schedules Using Simple Temporal Constraint Networks Stephen Smith, Research Professor, Carnegie Mellon University, 5000 Forbes Avenue, The Robotics Institute, Pittsburgh, PA, 15213, United States, [email protected], Anthony Gallagher, Terry Zimmerman, Laura Barbulescu, Zachary Rubinstein
We consider the problem of maintaining the schedules of a team of agents operating jointly in an uncertain execution environment. The objective is to maximize the quality of executed actions, and an action taken by a given agent may depend on those taken by others. Since no single agent has a complete view of the problem, agents must coordinate to identify scheduling decisions that maximize global quality. We describe a solution approach rooted in the use of Simple Temporal Network models.

2 - Writing a Basic Statistics Textbook Samuel Wathen, Iterim Dean, Operations, Coastal Carolina University, Wall College of Business, Conway, SC, 29528, United States, [email protected]
A basic statistics textbook is directed at two audiences: students and instructors. Writing for these audiences presents the challenge of presenting the material with many real world examples, illustrations, and scenarios so that students can connect with the topic of statistics. Instructors appreciate the readability of a text and the inclusion of a generally accepted set of topics. Instructors also desire many different forms of course support.

3 - Managerial Statistics: A Case-Based Approach Brett Saraniti, Hawaii Pacific University, 1060 Bishop Street, #402, Honolulu, HI, United States, [email protected]
The business statistics course at the Kellogg School of Management has been using a case-based approach for the past decade. The course packet has evolved into a new textbook by Thomson-Southwestern titled “Managerial Statistics: A Case-Based Approach.” Traditional topics are motivated by real world cases using a hands-on approach that utilizes a simple Excel macro.

I SA08
Technology Strategy: Entry, Adoption, and Acquisition
Sponsor: Technology Management Sponsored Session
Chair: Sebastian Fixson, Adjunct Assistant Research Scientist, Industrial and Operations Engineering, University of Michigan, Ann Arbor, MI, United States, [email protected] 1 - Evolution of a Technological System and Organizational Dynamics: A Subsystem Level Analysis Jaegul Lee, Postdoctoral Fellow, Carnegie Mellon University, Pittsburgh, PA, 15213, United States, [email protected]
This research examines the link between technological evolutions of the automobile emission control system at the subsystem level and the entry as well as exit patterns of firms’ innovation activities. Findings of this research suggests that firms’ entry and exit during the subsequent cycles of technological change is influenced significantly by the establishment of the subsequent technological regime and ecological processes of involved firm population.

2 - Systematic Local Search for Constraint Optimization Bill Havens, Simon Fraser University, School of Computing Science, Burnaby, BC, V5A1S6, Canada, [email protected]
We define a new search framework for solving constraint satisfaction problems (CSPs) and constraint optimization problems (COPs). The frameworks suggests a multiplicity of new hybrid search methods which combine aspects of stochastic local search, systematic constructive (bactrack) search algorithms and nogood learning techniques. Recently we have shown significant progress in using this hybrid method for hard random CSPs, job shop and open shop scheduling problems.

3 - A Survey of Existing and Unexplored Military Applications of Constraint Programming Jean-Paul Watson, Sandia National Laboratories, PO Box 5800, MS 1110, Albuquerque, NM, 87185-1110, United States, [email protected]
Industrial logistics problems, e.g., transportation and scheduling, have been used to showcase the power and flexibility of constraint programming (CP) technology. However, the transference to military logistics problems - which are often richer than their industrial counterparts - is limited. We survey prior successful applications of CP in military logistics, and detail a number of unexplored problems for which CP may be particularly suitable.

53

SA09

INFORMS PITTSBURGH — 2006
4 - Effects of Learning and Cooperation on New Firm Marketing John Angelis, Case Western Reserve University, 10900 Euclid Avenue, Cleveland, OH, United States, [email protected]
New firms face liabilities of newness. These liabilities exist both in a firm’s marketing ability and in customer retention and acceptance. We investigate how a firm’s optimal marketing expenditures should change due to learning over time and the firm’s endogenous cooperation level with other firms.

2 - Knowledge Relatedness and Industry Life Cycle: Evidence from the Early Automotive Airbag Industry Sebastian Fixson, Adjunct Assistant Research Scientist, Industrial and Operations Engineering, University of Michigan, Ann Arbor, MI, United States, [email protected], Wonhee Lee
Linking the concepts of knowledge relatedness and industry life cycle we explore in this paper the dynamic nature of knowledge relatedness by comparing industry pre-takeoff and post-takeoff stages of the automotive airbag industry. We find that while prior to industry take-off firms’ inward focus produces more valuable innovations, post industry take-off firms need to explore more across their firm boundaries to produce valuable innovations.

I SA11
Real-Life Sports Applications
Sponsor: OR in Sports Sponsored Session
Chair: Michael Trick, General Chair, INFORMS Anuual Meeting, Carnegie Mellon University, Tepper School of Business, 5000 Forbes Avenue, Pittsburgh, PA, 15217, United States, [email protected] 1 - Moneyball and the Mavs. Wayne Winston, Professor, Indiana University, 3453 Maritime Court, Bloomington, IN, 47401, United States, [email protected]
We will describe our player and lineup evaluation system that has been used for Dallas Mavericks for 6 seasons. Over that time period the Mavs won more regular season games than any other NBA team. We will pay particular attention to how the coaching staff uses our data.

3 - Technology Adoption and Performance Among Medical Service Providers J. Lamar Pierce, Carnegie Mellon University, Tepper School of Business, Pittsburgh, PA, 15213, United States, [email protected], Aaron Chatterji
When and how to acquire new medical equipment is an important strategic decision for health care firms. Especially for new firms, significant tradeoffs exist between financial stability and quality of services. This study uses data on medical equipment to investigate the impact of technology adoption and financing on firm performance.

4 - Dynamic Competitive Advantage Through Innovation: M&A, Cooperation, Contracting & Tech. Performance Giovanni Valentini, Bocconi University, ISEA, Milan, Italy, [email protected]
Recent evidence suggests that the innovation process is increasingly involving partners beyond a firm’s boundaries, including research companies, business partners, and universities. This work explores issues on the drivers and the consequences of an open innovation strategy.

I SA09
R&D Decisions in Entrepreneurial Settings
Cluster: New Product Development Invited Session
Chair: Nitin Joglekar, Associate Professor, Boston University, 525 Commonwealth Avenue, Boston, MA, 02215, United States, [email protected] 1 - Strategic Isomorphism, Legitimacy, and IPO Markets David Deeds, Associate Professor, The School of Management, University of Texas at Dallas, PO Box 830688, SM 43, Richardson, TX, 75083-0688, United States, [email protected], Theodore A. Khoury
Do investors reward conformity? In emerging industries, do they act as agents of isomorphism? We address these questions by examining the impact of a firm’s strategic conformity on the flow of resources into the firm using a sample of 150 biotech firms that issued an IPO between 1982 to 2001. Using IPO performance as a measure of constituent-based legitimacy, we find that firms who conform to specific strategies are rewarded but we also find areas in which firms are penalized for conformance.

2 - ChartMine: Using Decision Theory and Data Mining to Improve On-Field Baseball Decision Making Eric Bickel, Assistant Professor, Texas A&M University, 236B Zachry Engineering Center, 3131 TAMU, College Station, TX, 77843-3131, United States, [email protected]
Over 300 high schools and universities have used ChartMine®, including over 30% of NCAA D1, to improve pitching/hitting strategies. ESPN has used ChartMine to provide information to viewers during the College World Series. ChartMine is a software system that combines efficient data collection with a powerful database analysis tool, grounded on decision analytic theory. I will describe ChartMine and discuss some lessons learned from trying to make money by applying OR to sports decisions.

3 - Umpire Crew Scheduling at the U.S. Open Adam Farmer, Logistics Engineer, The Home Depot, 2455 Paces Ferry Road, Atlanta, GA, 30339, United States, [email protected], Jeffrey Smith
Scheduling umpires at the U.S. Open can require a significant amount of time due to several factors. The authors developed a system to automate the scheduling process and successfully implemented the system at the 2004 and 2005 U.S. Opens. In 2006, the system was enhanced to allow for the assignment of crew chiefs and incorporate ongoing assignment history throughout the course of the tournament. Results of the updated system’s implementation at the 2006 U.S. Open will be discussed.

2 - Diagnosing Unforeseeable Uncertainty in Startups Christoph Loch, Professor of Technology Management, INSEAD, Boulevard de Constance, Fontainebleau, France, [email protected]
Novel startups often face unforeseeable uncertainty, or influence factors that cannot be identified beforehand. However, management systems (e.g., whether to proceed with parallel trials or by probing and learning) must be set at the outset. This contradiction can be resolved by asking for gaps and inconsistencies in one’s knowledge: the PRESENCE of unforeseeable uncertainty can be diagnosed. The extent of uncertainty and complexity allow guidance in choosing the management systems.

4 - Scheduling Major League Baseball Umpires Tallys Yunes, University of Miami, Department of Management Science, Coral Gables, FL, 33124-8237, United States, [email protected], Michael Trick, Heather Barringer, Jonathan Levine
In this talk we address the problem of constructing schedules for MLB umpire crews. We will discuss a few heuristic approaches for this problem and present computational results on real data. Our work created the 2006 MLB umpire schedule.

I SA12
2006 Dantzig Dissertation Award Finalists
Cluster: Dantzig Dissertation Prize Invited Session
Chair: Robert Shumsky, Associate Professor, Tuck School, Dartmouth College, 100 Tuck Hall, Hanover, NH, 03755, United States, [email protected] 1 - Availability Management for Configure-to-Order Supply Chain Systems Ching-Hua Chen-Ritzo, Research Staff Member, IBM T.J. Watson Research Center, Mathematical Sciences, 1101 Kitchawan Road, Yorktown Heights, NY, 10598, [email protected]
We address the problem of aligning demand and supply in configure-to-order systems. Using stochastic programming methods, this study demonstrates the value of accounting for the uncertainty associated with how orders are configured, as well as the value of component supply flexibility in the presence

3 - Balancing R&D and Marketing Tension in Startup Firms Moren Levesque, Associate Professor, University of Waterloo, Waterloo, ON, Canada, [email protected], Nitin Joglekar
When should the owner of a startup firm acquire working capital and how should she distribute this capital between R&D - that improves product quality and marketing expenses - that builds the sales up? The industry focus is software, where both R&D and marketing are significant fractions of overall expenses. We develop a model to study optimal resource acquisition and allocation across successive stages of startup growth.

54

INFORMS PITTSBURGH— 2006
of order configuration uncertainty. We position our models within an increasingly popular business process called Sales and Operations Planning. Computational analyses were performed using data obtained from IBM System and Technology group.

SA15

2 - Maximizing the Efficiency of the U.S. Liver Allocation System through Region Design Nan Kong, Assistant Professor, Industrial and Management Systems Engineering, University of South Florida, ENC 2505, Tampa, FL, 33620, United States, [email protected]
Allocating organs for transplantation has been a contentious issue in the U.S. for decades. One of the remaining central questions is how to facilitate regional distribution of organ procurement organizations. We formulate this problem as a large-scale set-partitioning problem to find the most efficient set of regions. We also design a branch-and-price algorithm for the problem, in which we apply a decomposition-based heuristic and several valid inequalities to algorithm effectiveness.

3 - Bayesian Approaches to Model Robust and Model Discrimination Designs Vincent Agboto, Research Scientist, Battelle, 505 King Avenue, Columbus, OH, 43228, United States, [email protected], Christopher Nachtsheim
A criticism frequently associated with the use of many optimality criteria in industrial experiments is the specific reliance on a statistical model. One may respond by claiming that the true model is an approximation of an unknown element of a known set of models. We introduce a Bayesian design method that accounts for parameter estimation and prior probability on the set of models of interest. We also introduce some Bayesian model discrimination approaches. Useful examples will be given.

3 - Inventory Control in a Build-to-Order Environment Melda Ormeci, School of Industrial & Systems Engineering, Georgia Institute of Technology, Atlanta, GA, 30332-0205, [email protected]
We look at the challenges and solution strategies of employing build-to-order in the context of global supply. We introduce a new shipping policy, ‘’Ship-toAverage”, which prescribes sending a fixed quantity, based on the long term average forecast, with each shipment and making adjustments only if the inventory strays outside a prescribed range. We provide theoretical support for the observed performance of this type of policy by addressing a Brownian control problem whose objective is to find the right balance between holding costs and the costs for adjusting shipment sizes.

4 - Impacts of Inspection Policies on Quality Coordination in a Supply Chain Hong Wan, Assistant Professor, School of Industrial Engineering, Purdue University, 315 N. Grant Street, West Lafayette, IN, 47907, United States, [email protected], Xiaowei Xu
This paper studies how inspection policies affect product quality in a supply chain. A manufacturer purchases components from a supplier. If the manufacturer coordinates with the supplier on component quality, she should use the All-or-None inspection policy; Otherwise, she should inspect at least a proportion of her order, but still fails to achieve the first-best component quality level. Supply chain efficiencies can be restores with a warranty-sharing contract.

5 - A Posterior Preference Articulation Method to Dual Response Surface Optimization Injun Jeong, Senior Researcher, Strategic Management Research Team / ETRI, 161 Gajeong-Dong, Yuseong-Gu, Daejeon, 305-700, South Korea, [email protected], Sungsoo Han, Kwangjae Kim
In dual response surface optimization, the mean and the standard deviation responses are often in conflict. To obtain a satisfactory compromise in such a case, a decision maker (DM)’s preference information on the tradeoffs among the responses should be incorporated into the problem. This paper proposes a posterior preference articulation method, which generates a comprehensive set of efficient solutions and then enables the DM to choose the best one from the set systematically.

I SA13
Teaching Case Competition I
Sponsor: Education (INFORM-ED) Sponsored Session
Chair: Tasha R. Inniss, Assistant Professor of Mathematics, Spelman College, 350 Spelman Lane, SW, Box 320, Atlanta, GA, 30126, United States, [email protected] 1 - Presentations of Finalists #1 and #2 for the Seventh Annual INFORMS Case Competition Tasha R. Inniss, Assistant Professor of Mathematics, Spelman College, 350 Spelman Lane, SW, Box 320, Atlanta, GA, 30126, United States, [email protected]
The FOUR finalists for the 2006 INFORMS Case Competition will deliver final presentations of their teaching cases to a panel of pre-selected judges and the audience. All are welcome to attend and observe these presentations as well as ask questions of the finalists. At the end of the four presentations, the judges will select the winner of the Case Competition. The winners and runners-up will be announced and presented with plaques as well as cash awards at the INFORMED Annual Business Meeting.

I SA15
Laboratory Auctions Experiments
Cluster: Auctions and e-Commerce Invited Session
Chair: Elena Katok, Associate Professor, Penn State University, 465 Business Building, University Park, PA, 16802, United States, [email protected] 1 - Reverse Auctions with Social Identity Sherry Li, Assistant Professor, University of Texas - Dallas, 2610 N. Floyd Road, School of Economics, Richardson, TX, 75083, United States, [email protected], Ernan Haruvy, Kutsal Dogan
In reverse auctions where the buyer has latitude to select the winner, social preferences may influence winner selection as well as the price offer. In a laboratory study, we induce social identity with preferences over paintings. Subjects in the role of sellers repeatedly make price offers and buyers can select any offer. Our investigation aims to outline the impact of the induced social identities on price offers and matches.

I SA14
Optimization in Quality Control
Contributed Session
Chair: Injun Jeong, Senior Researcher, Strategic Management Research Team / ETRI, 161 Gajeong-Dong, Yuseong-Gu, Daejeon, 305-700, South Korea, [email protected] 1 - A Robust Design Optimization Model Considering Multiple Responses of Different Types Jamison Kovach, Assistant Professor, University of Houston, 312 Technology Building, University of Houston, Houston, TX, 77204, United States, [email protected], Byung Rae Cho
Several robust design optimization models have been proposed in the literature, yet few have specifically addressed problems involving the simultaneous consideration of multiple quality characteristics of different types. Here, we propose a new robust design optimization strategy which is a more flexible and structured approach for solving multiresponse problems effectively.

2 - Bidder Behavior in Multiple-Good Sealed Bid Auctions where the Number of Bidders is Unknown Svetlana Pevnitskaya, Assistant Professor of Economics, Florida State University, 288 Bellamy Building, Department of Economics, Tallahassee, FL, 32306-2180, United States, [email protected], Kurt Schnier, Mark Isaac
This paper analyzes the individual bidding data from a series of sealed-bid auctions with uncertainty over how many participants bidders will be bidding against. For risk-neutral individuals, there is a theory of bidding in this environment. In this paper we 1) Examine the performance of this theory using our individual bidding data, and 2) Extend the analysis to include the implications of risk preferences and potentially risk-averse bidders.

2 - A Computationally Efficient Robust M-Estimator Jonathan Stephenson, SAS Institute, 100 SAS Campus Drive, Cary, NC, 27513, United States, [email protected]
A bounded-influence estimator is considered and its asymptotic properties are discussed. The influence function is depicted and the breakdown point is explored. Algorithmic choices are outlined with application to linear and nonlinear parameter estimation. Extensions to solving linear and nonlinear systems.

3 - A Direct Test of Risk Aversion in First Price Sealed Bid Auctions Elena Katok, Associate Professor, Penn State University, 465 Business Building, University Park, PA, 16802, United States, [email protected], Richard Engelbrecht-Wiggans
We present theory and experiments designed to provide a direct test of risk aversion as the explanation of “overbidding” relative to the risk-neutral Nash equilibrium in sealed-bid first price (SBFP) auctions. We conclude that risk aversion fails the test - it cannot, by itself, explain overbidding in SBFP auctions.

55

SA16 I SA16

INFORMS PITTSBURGH — 2006
2 - Pricing and Ordering Policies When Demand is a Stochastic Function of Prices in Multiple Periods Mehmet Gumus, Department of Industrial Engineering & Operations Research, University of California, Berkeley, CA, 94720, United States, [email protected], Phil Kaminsky, Hyun-Soo Ahn
We consider a joint pricing and ordering problem where stochastic demand realized at each period is influenced by the current price as well as prices at previous periods. We formulate finite-horizon stochastic dynamic programs for a firm maximizing its profit. We characterize the structure of optimal policies for make-to-order and make-to-stock models and conduct preliminary computational study.

Dynamic Pricing and Inventory Control
Cluster: Dynamic Pricing and Forecasting Invited Session
Chair: Panos Kouvelis, Professor, Washington University in St. Louis, Olin School of Business, One Brookings Drive, Campus Box 1133, St. Louis, MO, 63130-4899, United States, [email protected] Co-Chair: Lingxiu Dong, Olin School of Business, Washington University in St. Louis, One Brookings Drive, Campus Box 1133, Saint Louis, MO, 63130, United States, [email protected] 1 - Flexible Capacity, Consumer Choice, and Competition Under Uncertainty Weiping Chen, Ph.D. Candidate, Department of Industrial and Systems Engineering, Virginia Tech, Blacksburg, VA, 24061-0118, United States, [email protected], Ebru Bish
We study the firm’s optimal capacity investment portfolio under a consumer choice model and responsive pricing. The firm produces two substitutable products and can invest in product-dedicated resources as well as a more expensive, flexible resource. We consider both monopoly and duopoly markets and characterize the structure of the optimal investment decision in both cases. We also investigate the benefits of flexible capacity under various situations and derive managerial insights.

3 - Structuring Long-Term Supply Contracts in the Presence of Spot Market Zhan Pang, PhD Student, The Chinese University of Hong Kong, The Department of SEEM of CUHK, R615, M. W. Mong Engineering Building, Hong Kong, China, [email protected], Youyi Feng
We consider structuring the long-term supply contract of single product between a make-to-stock manufacturing system and its institutional customer(s), both of which are able to trade the product in a spot market. We structure a contract that not only provides strategic value of mitigating variability of sale and cash inflow, but also increases net revenue. An optimal supply curve is used to reach a Nash equilibrium through a series of bilateral bargaining game so as to strike a win-win deal.

2 - Dynamic Allocation to Multiple-Class Demand with Continuous Review Joseph Milner, Professor, [email protected], Panos Kouvelis, Ding Qing
We consider how a firm should dynamically allocate inventory to different customer classes that are differentiated by price and willingness to wait for delayed delivery as well as reorder in a continuous review environment. We establish the optimal policy and introduce efficient and robust heuristics.

4 - Price Optimally and Leverage the Marketing Effort When Entering a Direct Market Gangshu Cai, Assistant Professor, Texas A&M International University, 311 Lake Carnegie, Laredo, TX, 78041, United States, [email protected], Sungjune Park, Moutaz Khouja
This work designs an aggregated discrete choice model enhanced with a marketing effort to identify the market share of the supplier and the retailer when they compete for the same market. We first analyze the multi-channel model with complete market information and provide optimal solutions to the scenarios before/after a Stackelberg game. We also study a multi-channel newsboy problem when considering with/without allowing the unsatisfied customers to switch from the retailer to the supplier.

3 - Dynamic Pricing of Seasonal Products in the Presence of Forward-Looking Consumers Yossi Aviv, Associate Professor of Operations Management, Olin School of Business, Washington University, Campus Box 1133, 1 Brookings Drive, St. Louis, MO, 63141, United States, [email protected], Amit Pazgal
We study the dynamic pricing of seasonal goods in the presence of strategic customers. Under limited inventory, consumers need to consider not only future prices, but the likelihood of stockouts, which depends on other customers’ behavior. We analyze the structure of optimal purchasing strategies, and the way in which strategic consumer behavior impacts pricing policies and expected revenue performance for the seller.

5 - Revenue Management Performance Drivers Carrie Crystal, PhD Student, Georgia Tech, College of Management, 800 W. Peachtree Street, Atlanta, GA, 30332, United States, [email protected], Mark Ferguson, Jeff Stratman
We present a theory-based model describing how key factors interact to impact revenue management performance and empirically test this theory. This study contributes to the literature by presenting a large scale comparison of practices and performance and indicating improvement areas for greatest impact.

4 - Dynamic Pricing and Inventory Control of Substitutable Products Zhongjun Tian, Olin School of Business, Washington University in St. Louis, Campus Box 1133, St. Louis, MO, 63130, United States, [email protected], Lingxiu Dong, Panos Kouvelis
We study the joint dynamic pricing and inventory control of multiple substitutable and perishable products for a retailer. The multinomial logit (MNL) model is adopted to capture consumer choice preferences. We derive the optimal dynamic pricing policy for the entire selling season, and identify key factors that affect the pricing guidelines. We study the effect of various pricing schemes on the initial inventory/assortment decisions and the interaction between the pricing and initial decisions.

I SA18
Investment Incentives and Risk Management in Electricity Markets
Sponsor: Energy, Natural Resources & The Environment Sponsored Session
Chair: Shmuel Oren, Professor, University of California, Berkeley, IEOR Department, 4119 Etcheverry Hall, Berkeley, CA, 94720, United States, [email protected] 1 - A Forward Capacity Market Model for Electricity Hung Po Chao, Director of Market Monitoring, Independent Systems Operator, New England, One Sullivan Road, Holyoke, MA, 01040, United States, [email protected]
We study a multi-period forward capacity market for electricity that includes a primary capacity auction and multiple secondary reconfiguration auctions. The capacity is traded as an option with an energy strike price, which effectively sets an offer cap, as the marginal revenue for the capacity from energy market is capped at that level. Characterization of dynamic equilibrium for energy and capacity markets is established in a network. Market power issue is investigated.

I SA17
Multiechelon Revenue Management Models
Contributed Session
Chair: Carrie Crystal, PhD Student, Georgia Tech, College of Management, 800 W. Peachtree Street, Atlanta, GA, 30332, United States, [email protected] 1 - Multiple Period Pricing Models Tao Huang, PROS Revenue Management, 3100 Main St. Suite 900, Houston, TX, 77002, United States, [email protected], Eylem Tekin
We consider pricing models in a multi-period setting. First, we look into the dynamic pricing problems when the initial service capacity is fixed and present closed-form solutions for single-product dynamic pricing problems. Second, we consider problems where both the price and the initial capacity are decision variables under stochastic demand. We formulate multi-stage stochastic programming models and derive structural properties of the optimal value functions.

2 - Free Allowances and Investments in a CO2 Constrained Restructured Electricity Market Yves Smeers, Professor, Universite Catholique de Louvain CORE/UCL, 34 voie du Roman Pays, Louvain La Neuve, Be, B-1348, Belgium, [email protected], Andreas Ehrenmann
We consider a stylised investment model of generation in a restructured electricity system with a CO2 cap and trade. We discuss the incentive to ivnestments arising from granting free allowances on the base of output or fuel consumption.

56

INFORMS PITTSBURGH— 2006
3 - Incentives of Generation Firms for Long-Term Investment in the U.S. Electricity Transmission System Enzo Sauma, Assistant Professor, Pontificia Universidad Catolica de Chile, Av. Vicuna Mackenna #4860, Raul Deves Hall, Piso 3, Macul, Santiago, Chile, [email protected], Shmuel Oren
We study the incentives that generation firms have in the U.S. electricity markets for supporting long-term social-welfare-improving transmission investments. We also examine how such incentives are affected by the ownership structure of financial transmission rights (FTRs). We show that generation firms may have disincentives for making socially beneficial transmission investments in the U.S. system. However, we also show how this situation could be reversed by correctly allocating FTRs.

SA21

I SA20
Communicating with Decision Makers
Sponsor: Decision Analysis Society Sponsored Session
Chair: James Felli, Sr. Research Advisor, Eli Lilly & Company, Lilly Corporate Center, Indianapolis, IN, United States, [email protected] 1 - Seven Principles for Effective Communication with Decision Makers Lawrence Phillips, Professor, London School of Economics & Political Science, Department of Operational Research, Houghton Street, London, NW3 1AH, United Kingdom, [email protected]
Applying seven principles will improve our communication with decision makers: (1) Understand the client’s problem, (2) Work as process consultants, (3) Engage with groups of key players rather than individuals, (4) Ensure meeting room arrangements facilitate interaction, (5) Use metaphor not jargon, (6) Model on-the-spot and display results graphically, and (7) Focus on creating aligned commitment.

4 - A Hybrid Bertrand-Cournot Model of Electricity Markets with Multiple Subnetworks Jian Yao, University of California, Department of IEOR, 4141 Etcheverry Hall, Berkeley, CA, 94720, United States, [email protected], Shmuel Oren, Benjamin Hobbs
We propose a hybrid Bertrand-Cournot model of electricity markets where inter-subnetwork connections are frequently congested but intra-subnetwork connections are not. In this model, generators behave Cournot regarding intersubnetwork transmission quantities, but Bertrand regarding intra-subnetwork transmission prices. Numerical examples suggest that this model leads to equilibrium prices higher than those from a homogeneous Bertrand model, but lower than those from a homogenous Cournot model.

I SA19
Forestry Applications I
Sponsor: Energy, Natural Resources & The Environment Sponsored Session
Chair: Susan King, U.S. Bureau of Labor Statistics, 2 Massachusetts Ave., NE, Room 3650, Washington, DC, United States, [email protected] 1 - Robust Optimization Applied to the Supply Chain of the Chilean Forest Industry Pamela Alvarez, Facultad de Ingenierìa, Departamento de Ciencias de la Ingenierìa, Universidad Andrés Bello, Chile, [email protected], Jorge Vera
The following work applies the methodology of robust optimization to problems of the supply chain of the Chilean forest industry in different planning levels. This methodology treats the uncertainty in optimization problems, associating with the uncertain problem a new one called robust counterpart. Specifically, robust models for problems of sawmill planning and problems of forest harvests are analyzed with uncertain yield parameters.

2 - Transparent Decision Analysis Robert Bordley, Technical Fellow, General Motors Research, 525 Choice Court, Troy, MI, 48084, United States, [email protected]
In most cases, the results, analyses and assumptions of a decision analysis should all be transparent to the stakeholders impacted by the decision. Much progress has been made on making the results of the analysis transparent. This paper presents practical ways of making the analysis visual.

3 - Metaphor Mapping James Felli, Sr. Research Advisor, Eli Lilly & Company, Lilly Corporate Center, Indianapolis, IN, United States, [email protected]
One obstacle faced by decision makers is a limited ability to visualize data. We present a new direction in data visualization: the use of metaphorical images to represent data sets. We will discuss the use of metaphor maps and their ability to succinctly convey information to a human observer.

I SA21
Decisions and Games Involving Outcomes Over Space and Time
Sponsor: Decision Analysis Society Sponsored Session
Chair: L. Robin Keller, Professor, University of California, Irvine, Merage School of Business, Irvine, CA, 92697-3125, United States, [email protected] 1 - Geographically-Oriented Preference Functions Craig Kirkwood, Professor, Arizona State University, W. P. Carey School of Business, PO Box 874706, Tempe, AZ, 85287-4706, United States, [email protected], L. Robin Keller
Some decisions involve alternatives with impacts that vary geographically. This paper applies multiattribute value and utility theory to develop functional forms for preference functions to address such decisions in a way that is both prescriptively sound and practical.

2 - Optimizing the Supply Chain for a Pulp Industry Using Decomposition Schemes Helene Gunnarsson, Linköping Institute of Technology, Division of Optimization, S-581 83, Linköping, Sweden, [email protected], Mikael Ronnqvist, Dick Carlsson
We consider the problem of planning the production, transportation and distribution of products in the pulp industry. The problem includes both binary and continuous decisions for production mix, terminal usage and ship routes, flows etc. We propose a mathematical model for the entire supply chain and we present a number of decomposition schemes to get a solution in acceptable computational time.

3 - Agent-Based Simulation of a Lumber Supply Chain Sophie D’Amours, FORAC Research Consortium, Université Laval, Pavillon Pouliot, Québec, Qc, G1K 7P4, Canada, [email protected], Luis Antonio Santa-Eulalia, Sébastien Lemieux, Jean-Marc Frayret, Alain Rousseau
We present an agent-based simulator to test supply chain configurations and planning approaches for the lumber industry. The architecture of the simulator and a proposed simulation framework are discussed. Impacts of different planning approaches are evaluated. The simulator builds on FIPA standards and integrates optimization modules. The application deals with sawmills, drying and finishing facilities, warehouses and customer buying behavior.

2 - Towards a General Preference Model Over Time Itself Jeffery L. Guyse, Associate Professor, California State Polytechnic University, Pomona, College of Business Administration, 3801 West Temple Avenue, Pomona, CA, 91768, United States, [email protected]
Trading monetary outcomes off over time has received significant attention by researchers. Likewise the psychological literature on perceived time is well developed. In this study we investigate the value function of time itself, independent of direct monetary considerations. The model can shed insights on how time is valued in a waiting or leisure setting, with special attention given to expectation and the shape of the value function.

4 - Collaborative Planning in Forest Transportation Mikael Ronnqvist, Norwegian School of Economics and Business Administration, Dept. of Finance and Management Science, NHH, Helleveien 30, NO-5045, Bergen, Norway, [email protected]
We describe and analyze how companies can coordinate their transportation planning and make a fair cost allocation or distribution of the cost savings. We make use of several economic models, including the Shapley value. We test the approach on a case study involving eight companies

57

SA22

INFORMS PITTSBURGH — 2006
The goal of this project is to develop a hierarchical mathematical program for the Mission and Means Framework. Based on scenario analysis, we propose a optimization model whose purpose is to construct a force that is robust across several conflicts and theatres for a given time horizon and against multiple threats which may or may not undergo modernization/adaptation. The mathematical program will optimize force survivability.

3 - Prostate Cancer Decision Analyst: Choosing the Best Treatment Jay Simon, Doctoral Student, University of California, Irvine, Operations & Decision Technologies, Irvine, CA, 92697-3125, United States, [email protected]
A man diagnosed with prostate cancer faces a difficult decision. Treatments have varying cure rates and a wide range of side effects. We develop a multipleobjective decision model under uncertainty, which will be shaped by the preferences and personal characteristics of the individual. This quantitative process will incorporate both user input and medical data, and will allow an individual prostate cancer patient to meaningfully compare treatments.

I SA23
Business Applications II
Contributed Session
Chair: Jason Niggley, University of Southern California, Marshall School of Business, 400C Bridge Hall, Los Angeles, CA, 90089, United States, [email protected] 1 - A Study of Corporate Cultural Differences Between Russia and the United States Arthur Rasher, Associate Professor of Management, Unniversity of Tulsa, 600 South College Avenue, Tulsa, OK, 74104, United States, [email protected], Galina Sokolova
The study compares the results of two studies of employment experiences in Moscow, Russia in 2004 and the same corporation in Tulsa, Oklahoma, United States, in 2005-2006. The studies used a qualitative research method - an adaptation of the critical incident method developed by Flanagan (1954). A record of incidents (a number of specific observations of particular differences) within two offices of an international company were collected and analyzed.

4 - Reputation, Trust, and Rebates: How Online Auction Markets can Improve their Feedback Mechanisms Lingfang Li, University of California, Irvine, 6336 Adobe Circle, Irvine, CA, 92617, United States, [email protected]
In current online auction markets only half of the buyers leave feedback after transactions, and nearly all of it is positive. This paper proposes a mechanism whereby sellers can provide rebates to buyers contingent on buyers’ providing reports. In both a pure adverse selection model, and a model with adverse selection and moral hazard, there exists a pooling equilibrium where both good and bad sellers choose the rebate option, even though their true types are revealed through feedback.

I SA22
Missions and Means Framework: Mathematical Perspectives
Sponsor: Military Applications Sponsored Session
Chair: Paul Tanenbaum, Director, Army Research Laboratory, Survivability/Lethality Analysis Directorate, ATTN: AMSRD-ARL-SL, Aberdeen Proving Ground, MD, 21005, United States, [email protected] 1 - A Mathematical Missions and Means Framework: Order-Theoretic Aspects Jonathan Farley, Science Fellow, Center for International Security and Cooperation, Stanford University, 616 Serra Street, Stanford, CA, 94305, United States, [email protected]
Order-theoretic aspects of the military missions and means framework are discussed.

2 - Applying Sequence Theory to Service Operations Jason Niggley, University of Southern California, Marshall School of Business, 400C Bridge Hall, Los Angeles, CA, 90089, United States, [email protected]
Starting with a theory of memory, how are related events remembered depending on the sequence in which they occur? Based on these findings, an optimal design of service operations exists.

I SA24
Perspectives on the Journal
Sponsor: Organization Science Sponsored Session
Chair: Linda Argote, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected] 1 - Perspectives on the Journal Linda Argote, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected]
The session will focus on INFORMS journal, Organization Science’s Current Status and Future Directions. The session will provide an overview of Organization Science, including its editorial team, the vision for the journal, objectives, and its operation. Statistics about review time and acceptance rates will be presented. Future directions including special issues will be described.

2 - Reflexive Control in Groups Vladimir Lefebvre, Researcher, Institute for Mathematical Behavioral Sciences, University of California, Irvine, 5100 Social Sciences Plaza, Irvine, CA, 92697, United States, [email protected]
A theoretical model of group behavior is introduced in this paper. We assume that each agent possesses a system of self-reflexion: a set of images of the self which can have sets of images of the self, etc. The model allows us to predict the agent’s state based on his system of self-reflexion and reflexive control over him conducted by other agents.

3 - Missions and Means via Formal Concept Analysis Stefan Schmidt, CEO, Phoenix Mathematics, Inc., 4045 Tellbrook Road, Las Cruces, NM, 88011, United States, [email protected]
In a domain of actions and communications, we investigate how constraints on capabilities influence the ability to achieve certain tasks. Our mathematical method of choice is Formal Concept Analysis: We interpret tasks as objects and capabilities as attributes; we then derive hidden dependencies between capabilities via the lattice of formal concepts. Finally, we extend our model using Reflexive Theory to demonstrate how wrongly perceived constraints may influence the decision making process.

2 - Aiming for a Successful Experience Richard Burton, Duke University, Fuqua School of Business, 1 Towerview Drive, Durham, NC, 27708, [email protected]
The session will discuss how to increase the likelihood of a successful experience when submitting a manuscript to Organization Science. Characteristics of manuscripts that editors and reviewers regard positively will be identified. Common shortcomings of manuscripts will be described. How to respond successfully to editor’s and reviewers’ comments will be discussed.

I SA25
Panel Discussion: OR Entrepreneurship #1
Cluster: OR Entrepreneurship Invited Session
Chair: Bruce Golden, University of Maryland, R.H. Smith School of Business, College Park, MD, 20742, United States, [email protected] 1 - Panel Discussion: OR Entrepreneurship Moderator: Bruce Golden, University of Maryland, R.H. Smith School of Business, College Park, MD, 20742, United States, [email protected], Panelists: Jean-Marc Rousseau, Yossi Sheffi, Sridhar Tayur, Irv Lustig, Michael Hooban
Many members of INFORMS have worked in an entrepreneurial capacity. Some have started and/or grown companies on a full-time basis. Others (e.g., professors) have started companies on a part-time basis. The purpose of this session is to tell some of the stories associated with these activities and to discuss common experiences, obstacles, and strategies.

4 - Closure Systems and State Spaces Alex Pogel, mathematician, Physical Science Laboratory, New Mexico State University, Las Cruces, NM, 88003, United States, [email protected]
In this talk we will discuss the use of a closure system (a mathematical structure corresponding to set systems closed under intersection) as data-driven state space formed from a two-dimensional description of a system. We will describe connections of closure systems to Formal Concept Analysis (Ganter, Wille, Wolff, et al), knowledge spaces (Falmagne, Doignon, et al), and MMF (Deitz, Sheehan, et al).

5 - Scenario Analysis: A Tool for Enhancing the Mission Means Framework Philipp Djang, Army Research Laboratory, Survivability/Lethality Analysis Directorate, ATTN:: AMSRD-ARL-SL-EI, White Sands Missile Range, NM, 88002, United States, [email protected]

58

INFORMS PITTSBURGH— 2006 I SA26
Data Mining Applications in HealthCare
Sponsor: Data Mining Sponsored Session
Chair: Shital Shah, Assistant Professor, Rush University Medical Center, 1700 W. Van Buren Street, 126B, Chicago, IL, 60612, United States, [email protected] 1 - Facilitating Performance Improvement at Academic Medical Centers - There’s Data Everywhere Sam Hohmann, University HealthSystem Consortium, 2001 Spring Road, Suite 700, Oak Brook, IL, United States, [email protected]
The University HealthSystem Consortium is an association of academic medical centers that gathers clinical and operational data from its members. The data is used to facilitate members’ performance improvement efforts. By integrating information, UHC has been able to combine various data sets to enhance AMCs’ self understanding of their performance and propel them in their improvement of quality, safety, and productivity. Examples will be shared.

SA28

3 - Heuristics for Flexible Resource Allocation in a Serial Line with Reliability Considerations Chen-hung Wu, Assistant Professor, National Tsing Hua University, Hsinchu, Taiwan, 101, Section 2 Kuang Fu Road, Hsinchu, TW, 30013, Taiwan, [email protected], Douglas Down, Mark Lewis
We consider the allocation of flexible resources in a serial line with server failures. In addition to dedicated servers at each station, the system is equipped with a fixed number of flexible servers that are available to be assigned to any station. We show in the two-station case that transition monotone optimal policies minimize average inventory holding costs. We discuss heuristics based on the two-server model that reduce average holding costs significantly.

4 - A Predictive Maintenance Policy Based on Two Explicative Variables Estelle Deloux, Ph.D. Student, Ecole des Mines de Nantes/IRCCyN, Rue Alfred Kastler, Nantes, 44307, France, [email protected], Christophe Berenguer, B. Castanier
We are interested in the modelling and the evaluation of a maintenance policy for a single-unit system with a single failure mode function of several measurable quantities. The failure mode can be partially explained by two variables stochastically increasing with time. The probability of failure is assumed to increase with these variables. Different maintenance policies will be discussed in regard with their performance when the associated decision parameters minimize a given cost criterion.

2 - Bayesian Networks for Employable Medical Expert Systems Rich Neapolitan, Professor, Northeastern Illinois University, 5500 N. St. Louis, Chicago, IL, 60515, United States, [email protected]
Medical expert systems date back to the 1970’s when MYCIN was created. In the 1990’s a number of Bayesian network-based medical expert systems were developed. However, virtually none of them are deployed. I review the Bayesian network and its use in medical expert systems. Then I discuss design considerations necessary if we want do create a system that can seamlessly be incorporated into medical practice. Finally, I present Promedas, which is a system designed according to these considerations.

I SA28
Multiscale Wavelets Analysis and Applications
Sponsor: Quality, Statistics and Reliability Sponsored Session
Chair: Judy Jin, Associate Professor, Department of Industrial and Operations Engineering, University of Michigan, IOE 2855, Ann Arbor, MI, 48109, United States, [email protected] 1 - Multiscale Methodology and Detectability Xiaoming Huo, Assistant Professor, [email protected]
Multiscale methodology leads to powerful processing tools, such as wavelets. It has been proven to be effective in determining boundaries in some detectability problems, which is a fundamental problem in detection and ubiquitous in engineering. The existing results give the asymptotic rate of the boundaries. More precise distributional results can be derived, and they reveal more accurate properties of statistics that are used. Some recent results and applications will be discussed.

3 - Relabeling Algorithm for Improving Prediction Quality Shital Shah, Assistant Professor, Rush University Medical Center, 1700 W. Van Buren Street, 126B, Chicago, IL, 60612, United States, [email protected], Andrew Kusiak
The relabeling algorithm (for retrieval of noisy instances with binary outcomes) iteratively retrieves, selects, and re-labels data instances to improve prediction quality. It emphasizes knowledge generalization and confidence rather than classification accuracy. The proposed approach was successfully tested on the standard benchmark Pima Indians Diabetes dataset as well as bladder cancer immunotherapy data.

I SA27
Maintenance for Dynamic Environments
Sponsor: Quality, Statistics and Reliability Sponsored Session
Chair: Julie Simmons Ivy, Assistant Professor, University of Michigan, 701 Tappan Street, Ann Arbor, MI, 48109, United States, [email protected] 1 - Integrated Decision Making on Reconfiguration and ConditionBased Maintenance (CBM) Jing Zhou, Ph.D. Candidate, University of Michigan, 2350 Hayward Street, Ann Arbor, MI, 48105, United States, [email protected], Julie Simmons Ivy, Dragan Djurdjanovic, Jun Ni
In a multi-unit load sharing system, the dynamic interactions between component deterioration, working load, and production yield necessitate an integrated decision making on reconfiguration and CBM. An integrated model for the joint optimization of the economic load allocation and the optimal maintenance level is developed and illustrated through a numerical example. The results demonstrate that the integrated decision policy outperforms the fragmented one in terms of reduced total cost.

2 - Wavelet Features from Wireless Vibration Sensor Array for Container Integrity Monitoring Satish Bukkapatnam, Associate Professor, School of Industrial Engineering & Management, Oklahoma State University, 322 EN, Stillwater, OK, 74078, United States, [email protected]
Of late, security and integrity of transportation container has assumed significant importance. Hurst exponents computed from wavelet analysis of vibration signals are found to capture the underlying burst patterns. The various integrity measures of a container, such as stability of the contents, terrain characteristics and transportation speed are estimated through the use of Hurst exponent and entropy features extracted through wavelet analysis of vibration signals.

3 - A Methodology for Classification and Detection of Helicopter Bearing Faults David He, Associate Professor, University of Illinois at Chicago, [email protected]
In this talk, the development of a general methodology using wavelet analysis integrated with data mining approaches for classification and detection of helicopter swashplate bearing faults will be reported. A successful application of the developed methodology to recent seeded fault tests for US Army CH-47D and MH-47E swashplate bearings will be discussed.

2 - Dynamic Modeling of Maintenance Policies Using Stochastic Petri Nets with Aging Tokens Vitali Volovoi, Senior Research Engineer, Georgia Institute of Technology, 270 Ferst Drive, Atlanta, GA, 30332-0150, United States, [email protected]
The benefits of improving existing maintenance policies are vast, as are the associated challenges. Appropriate resource allocation is impossible without comprehensive system-level modeling of failure and repair. Dynamic interactions among factors affecting operations, degradation, and maintenance must be quantitatively accounted for. To this end, a framework based on Petri nets is introduced; it is capable of modelling imperfect inspection and maintenance, opportunistic repair, and logistics.

4 - Multiscale Feature Extraction for Embedded Operation Monitoring and Diagnosis Judy Jin, Associate Professor, Department of Industrial and Operations Engineering, University of Michigan, IOE 2855, Ann Arbor, MI, 48109, United States, [email protected], Gabrielle Tremblay, Yu Zhou
For a complex system consisting of multiple embedded operations, it is challenging to monitor and diagnose individual operation conditions based on the system sensing signals. This research will present a systematic way how to use multiscale wavelets analysis to maximize the diagnosability of extracted features for individual operations in both time and frequency domains.

59

SA29 I SA29

INFORMS PITTSBURGH — 2006
2 - The Effect of Electronic Medical Records on Hospital Performance Abby Swanson, Assistant Professor, Department of Health Administration & Policy, Medical University of South Carolina, 151 Rutledge Ave., Building B, PO Box 250961, Charleston, SC, 29425, [email protected], Yasar A. Ozcan
This study examines the prevalence and performance of Electronic Medical Records (EMRs) in acute care hospitals. Using logistic regression, organizational and environmental factors associated with hospital EMR use are identified. Data Envelopment Analysis and post hoc evaluations performed for comparison of hospitals with and without EMRs in efficiency and quality.

Resource Allocation and Facility Location in Healthcare
Cluster: Joint Cluster Healthcare/ HAS: Healthcare Engineering Invited Session
Chair: Julie Swann, Assistant Professor, Georgia Institute of Technology, Industrial and Systems Engineering, Atlanta, GA, 30332, United States, [email protected] Co-Chair: Christina Scherrer, Assistant Professor, Southern Polytechnic State University, 1100 S. Marietta Parkway, Atlanta, GA, 30060, United States, [email protected] 1 - Emergency Response Planning Eva Lee, Associate Professor, School of Industrial and Systems Engineering, Georgia Institute of Technology, Atlanta, GA, 30332, [email protected], Yang Zhang
We describe our work with state emergency response departments on facility location and planning for natural disasters, infectious disease outbreaks, and bioterrorism attacks. Optimization models and large-scale simulation will be described, and an actual field exercise related to an Anthrax drill will be reported.

3 - Deciding What Time Patients Should be Ready on the Day of Surgery without Procedure-Specific Data Ruth E. Wachtel, Associate Professor, Department of Anesthesia, University of Iowa, Iowa City, IA, 52242, [email protected], Franklin Dexter
No hospital has reported successfully implementing consensus guidelines for reduced NPO times. A patient can drink to 12 noon for surgery at 2 PM. Yet, 2 PM is the actual start time, not scheduled. For one preceding case, the start time of the next case follows a 3-parameter log normal distribution centered at the scheduled duration of the preceding case. Multiple previous cases are treated as one virtual case. Nurses access www.CaseDuration.com for distribution-free 0.05 prediction bounds.

2 - Network-Based Facility Location for Community Health Centers Christina Scherrer, Assistant Professor, Southern Polytechnic State University, 1100 S. Marietta Parkway, Atlanta, GA, 30060, United States, [email protected], Paul Griffin, Julie Swann
We determine the best location and medical service offerings for Community Health Centers (CHCs) using a network-based facility location approach. We expand upon an earlier model to consider weighted disease objectives and compare locations chosen by our model to existing locations based on several public health measures. We also use the model to analyze different healthcare access and coverage policy mechanisms for cost and effectiveness.

4 - Hospital Pediatric Specialization: Implications for the Delivery of Patient Care Mitch Harris, Ph.D. Candidate, NACHRI, 401 Wythe Street, Alexandria, VA, 22314, [email protected], Yasar A. Ozcan
Policy makers, payors and patients have shown increased interest in the concept of hospital specialization and its implications on effectiveness and quality of care. We present a model with levels of care complexity and hospital specialization that must be in fit for outcomes to be maximized.

3 - The Davao City Health System: An Approach to Optimally Locating Community Health Facilities Michael Johnson, Extern, Carnegie Mellon University, Center for Economic Development, 5000 Forbes Avenue, Pittsburgh, PA, 15213-3890, United States, [email protected], Myra Esmeralda Silva
Davao City, Philippines operates a two-tiered health system in which a large network of community-based health centers feeds into a much smaller network of regional hospitals. There is evidence of substantial inequities and gaps in service coverage. We develop an alternative two-tier system, in which a lower tier of inexpensive health outpostsis links to a smaller number of health centers. We present computational results for this hierarchical facility location problem and policy implications.

I SA31
Financial Engineering
Cluster: Financial Engineering and Risk Management Invited Session
Chair: Kay Giesecke, Assistant Professor, Management Science & Engineering, Stanford University, Stanford, CA, 94305, United States, [email protected] 1 - Connections Between Singular Control and Optimal Switching Pascal Tomecek, PhD Student, School of OR&IE, 294 Rhodes Hall, Cornell University, Ithaca, NY, 14853, United States, [email protected], Xin Guo
We explore new connections between finite variation singular control and sequential optimal stopping problems. We apply our results, along with the standard probabilistic approach to singular control, to extend the theory of reversible investment. We give sufficient conditions on the existence of an optimal solution and on the regularity of the value function. Application areas include energy and credit risk management, and markets with friction.

4 - Supply Chain Coordination and Influenza Vaccination (Submitted for the 2006 Pierskalla Award) Hamed Mamani, PhD Student, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Building E40-130, Cambridge, MA, 02139, United States, [email protected], David Simchi-Levi, Stephen Chick
Billions of dollars are being allocated for influenza pandemic preparedness. Vaccination is a primary weapon for fighting influenza outbreaks. Differences that distinguish the influenza vaccine supply chain from the typical supply chains include a nonlinear value of sales and vaccine production yield issues. We show that production risks, taken by the manufacturer, lead to an insufficient supply of vaccine and design a variant of the cost sharing contract that can coordinate the supply chain.

2 - Portfolio Choice Under Model Uncertainty Stefan Weber, Assistant Professor, School of OR&IE, Cornell University, 279 Rhodes Hall, Ithaca, NY, 14853, United States, [email protected]
We consider the problem of utility maximization under model uncertainty in the presence of both cost and risk constraints. Downside risk is measured by utilitybased shortfall risk. We assume that investors take a worst case approach. By means of its dual problem, the optimization problem can explicitly be solved.

I SA30
Contemporary Operational and Strategic Issues in Hospitals
Sponsor: Health Applications Section Sponsored Session
Chair: Yasar A. Ozcanm , Virginia Commonwealth University, PO Box 980203, 1008 E. Clay Street, Richmond, VA, 23298-0203, [email protected] 1 - Developing Fair and Feasible Schedules for Residents On-Call Sarah Root, University of Michigan, 1205 Beal Avenue, Ann Arbor, MI, United States, [email protected], Amy Cohn
Scheduling medical resident call shifts is challenging because of the many complex, competing feasibility rules to be satisfied. Furthermore, finding a feasible schedule is not enough, because it may be very imbalanced across residents. We present a scheduling approach that incorporates both feasibility and “fairness” and show real-world computational results.

3 - Cross Market Calibration of Credit Spreads Rajnish Kamat, Vice President, MSCI Barra, 2100 Milvia Street, Berkeley, CA, 94704, United States, [email protected], Lisa Goldberg, Vijay Poduri
We use Credit Default Swap spread data along with equity market and balance sheet information to build a next generation structural model. The model uses an incomplete information framework and provides greater flexibility in calibration compared to the Merton model. Applications include rich-cheap analysis of credits, credit portfolio monitoring, pricing less liquid names, and deriving exposures to market variables for hedging.

60

INFORMS PITTSBURGH— 2006 I SA32
Best Financial Services Student Paper Competition
Sponsor: Financial Services Sponsored Session
Chair: Suresh Nair, University of Connecticut, Storrs, CT, 06268, [email protected] 1 - Credit Spreads, Optimal Capital Structure, and Implied Volatility Nan Chen, Assistant Professor, The Chinese University of Hong Kong, Room 709A, William M. W. Mong Engineering Building, Shatin, N. T., Hong Kong, [email protected]
We propose a two-sided jump model for credit risk by extending the Leland-Toft endogenous default model based on the geometric Brownian motion. The model shows that jump risk and endogenous default can have significant impacts on credit spreads, optimal capital structure, and implied volatility of equity options. Pricing formulae of credit default swaps and equity default swaps are also given.

SA34

3 - Many Fields of Battle: How Cost Structures Affect Competition Across Multiple Markets Li Zou, Ph. D. Candidate, University of Maryland, Robert H. Smith School of Business, College Park, MD, United States, [email protected], Martin Dresner, Robert J. Windle
Traditional analysis of multipoint competition suggests that firms compete less intensely with one another when they have strategic contacts across multiple markets (the mutual forbearance effect). However, this argument does not account for the moderating role that production cost plays in these market relationships. This article examines the collusion-enhancing effects of multimarket contact and its effect on competition by using a model verified through data from the U.S. airline industry.

2 - Discrete-Time Optimization of Consumption and Investment Decisions Amit Bhandari, PhD Candidate, Industrial Engineering & Management Sciences, Northwestern University, 2145 N. Sheridan Road, Evanston, IL, 60208, United States, [email protected]
We extend Samuelson’s (1969) discrete-time dynamic consumption and investment optimization problem to the case where the investor is intolerant of any decline in her standard of living. This is interpreted as a constraint that the consumption rate is non-decreasing over time. The investor first guarantees a consumption perpetuity at the current consumption rate and then allocates the remaining wealth under a state-dependent, adjusted coefficient of relative risk aversion.

4 - Resource Allocation to a Portfolio of Modernization Technologies for the National Airspace System Senay Solak, School of Industrial and Systems Engineering, Georgia Institute of Technology, 765 Ferst Drive, Atlanta, GA, 30332, United States, [email protected], Paul Collopy, Ellis Johnson, John-Paul Clarke
We develop common performance assessment measures and analyze correlation levels for a portfolio of National Airspace System modernization technologies. We then propose a portfolio optimization model to determine the optimal allocation of available resources to the technologies in consideration. In addition, the computational difficulty of the proposed model is analyzed, and appropriate solution methods are described.

I SA34
Games of Cooperation and Competition in Transportation
Sponsor: Transportation Science & Logistics Sponsored Session
Chair: Richa Agarwal, PhD Candidate, School of Industrial and Systems Engineering, Georgia Institute of Technology, [email protected] 1 - Mechanism Design for Sustainable Alliances Among Sea Carriers Richa Agarwal, PhD Candidate, School of Industrial and Systems Engineering, Georgia Institute of Technology, [email protected], Özlem Ergun
In transportation industry, alliances among carriers are an important factor in staying competitive. Though working in close liaisons, carriers are governed by their own self-interests. We present membership mechanisms, by adopting a rationality based approach using notions from game theory and inverse optimization, to allocate resources and benefits among carriers for forming sustainable alliances.

3 - Pricing and Hedging of Volatility Derivatives Ashish Jain, Columbia Business School, 3022 Broadway Uris Hall, New York, NY, 10027, United States, [email protected]
We propose a methodology to price and hedge volatility derivatives. We derive a no arbitrage relationship between variance swaps and other volatility derivatives. Using this argument, we derive a partial differential equation to price volatility swaps and variance options in the Heston stochastic volatility model and compute the volatility and variance strikes using numerical transform inversion techniques and simulation.

I SA33
Modeling the National Airspace System (NAS)
Sponsor: Aviation Applications Sponsored Session
Chair: John Gulding, Senior Strategist, Federal Aviation Administration, 800 Independence Avenue SW, Washington, DC, 20591, United States, [email protected] 1 - Benefit Estimation in Presence of Demand Uncertainty Dipasis Bhadra, Principal Economist, MITRE/CAASD, 7515 Colshire Avenue, McLean, VA, 22102, United States, [email protected], Dave Knorr
Estimation of future benefits from federal programs designed to improve air transportation infrastructure using deterministic or point forecasts could be misleading in presence of uncertainty. Uncertainty rooted in airlines decisions, for example, make the forecasts at the airports very difficult, especially in the short run. We offer a framework where demand uncertainties can be incorporated in benefit estimation. We illustrate this framework by using an FAA investment program.

2 - Network Games with Atomic Players Jose Correa, Assistant Professor, Universidad Adolfo Ibáñez, Av. Presidente Errazuriz 3485, Las Condes, Santiago, Chile, [email protected], Roberto Cominetti, Nicolas Stier-Moses
We study network and congestion games with atomic players that can split their flow. As there can be players controlling a significant fraction of the flow, these games readily applies to competition among freight companies and telecommunication network service providers. We analyze the worst-case inefficiency of Nash equilibria and conclude although self-interested competitors will not in general achieve a fully efficient solution, the loss is not too severe.

2 - A Characterization of Days of the Week in the National Airspace System Brendan Hogan, Sr. Simulation Modeling Engineer, MITRE/CAASD & University of Virginia, 7515 Colshire Avenue, McLean, VA, 22102, United States, [email protected]
The NAS consists of complex interactions between aircraft operators and the ATC system handling the flights. To gain insight into these patterns of activity with respect to day of the week, we use principal component analysis to demonstrate similarities between individual days. This information is helpful in selecting representative traffic days for the purpose of simulation or other analyses. In addition this technique can be extended to show trends over time and across user groups.

3 - The Impact of Pricing and Buy-Back Menus on Supply Chain Performance Pranava Goundan, Doctoral Candidate, Operations Research Center, MIT, 77 Massachusetts Avenue, Building E40-149, Cambridge, MA, 02139, United States, [email protected], Ann Chan, David Simchi-Levi
We study the impact of a Buy-back Menu contract in a single-supplier, multiretailer supply chain facing linear additive demand, that makes both pricing and inventory decisions. We compare two systems: competing and coordinating retailers, and show that using the proposed contract in both systems, the supplier may improve the performance of the supply chain to at least 75%, while still being guaranteed at least 50% of the optimal global supply chain profit.

61

SA35 I SA35

INFORMS PITTSBURGH — 2006
3 - Greedy Rules vs. Real-Time Complete Reoptimization Strategies for the Online Inventory Routing Hani Mahmassani, University of Maryland, 3128 J. H. Kim Building, College Park, MD, 20742, United States, [email protected], Ricardo Giesen, Patrick Jaillet
We study strategies for the online inventory routing problem (OIRP) under realtime information. We propose a greedy heuristic and a mixed-integer programming (MIP) formulation for the off-line version and introduce different rolling horizon strategies for the OIRP. We examine potential benefits of updating plans based on real-time information.

Theoretical and Practical Applications of Integrating Demand and Network Models
Sponsor: Transportation Science & Logistics Sponsored Session
Chair: Laurie Garrow, Assistant Professor, School of Civil and Environmental Engineering, Georgia Institute of Technology, 790 Atlantic Drive, NW, Atlanta, GA, 30332, United States, [email protected] 1 - Theoretical and Practical Aspects of Modeling Network Equilibrium with Road Pricing Peter Vovsha, Principal, PB Consult, 5 Penn Plaza, 19th Floor, New York, NY, 10001, United States, [email protected]
The current paper is intended to systematically analyze possible operational approaches to a more consistent and theoretically sound modeling of road pricing. It starts with a formulation of a “seed” equilibrium model, discuss some basic aspects and pitfalls associated with practical implementation of this model, and then gradually adds features and discuss possible enhancements.

I SA37
Economics of Information Systems
Sponsor: Information Systems Society Sponsored Session
Chair: Indranil Bardhan, Assistant Professor, The University of Texas at Dallas, School of Management, SM 41, 2601 N. Floyd Road, Richardson, TX, 75083-0688, United States, [email protected] 1 - Testing a Theory of Network-Based Knowledge Spillovers in the Agglomeration of Technology Industries Ajay Kumar, Doctoral Student, The University of Minnesota, Carlson School of Management, Minneapolis, 55455, United States, [email protected], Robert Kauffman
We analyze the impacts of IT on the agglomeration of technology industries based on the multiple theoretical perspectives. We show that increasing use of IT diminishes the impetus of both market size linkages and congestion. It also leads to new IT industry agglomerations around thick labor markets. Agglomerated industries benefit from localized and network-based knowledge spillovers. The net impact of these effects leads to greater IT industry agglomeration.

2 - Integrated Activity-Based Modeling and Dynamic Traffic Micro-Assignment Simulator Travis Waller, University of Texas at Austin, 1 University Station C1761, Austin, TX, United States, [email protected], Chandra Bhat
This talk will discuss research related to an integrated Activity-Based/DTA model (CEMDAP and VISTA). Given the disaggregate data on socio-demographics, activity-travel environment and prescribed policy actions for the base year, CEMDAP generates detailed activity-travel patterns for each individual in a population which serve as inputs for the integrated DTA model. Technical, computational, and practical issues involved in this demand-supply integration problem will be discussed.

3 - Dulles Metrorail - Congestion Mitigation Plan Microsimulation John Duesing, Senior Associate, Cambridge Systematics, 405 Lexington Avenue, 26th Floor, New York, NY, 10174, United States, [email protected]
Cambridge Systematics, as the developer of the Tyson’s Corner Travel Demand Model for the County of Fairfax, developed a methodology for assessing the diversions and re-routing that will result from the construction of the Metrorail Orange line. We proposed integration of the Tyson’s Corner demand model with a VISSIM traffic microsimulation model of the study area to adequately evaluate the traffic impacts in the corridor. This integration will allow for modal shifts, time period shifts, as well as route shifts that will occur due to construction impacts.

2 - A Case-Based Model of Intellectual Property Reuse Barrie R. Nault, Professor and Chair, University of Calgary, Haskayne School of Business, Calgary, AB, T2N 1N4, Canada, [email protected], John D. Mitchell, Victoria L. Mitchell
Based on cases from the semiconductor industry, we build a model of intellectual property (e.g., software) reuse. We consider designer skills, a reuse repository, and the role of quality standards and incentives. Our results highlight a tradeoff between quality standards motivating reuse while reducing incentives for building new repository contributions.

I SA36
Dynamic Problems in Rail and Road Transportation
Sponsor: Transportation Science & Logistics Sponsored Session
Chair: Huseyin Topaloglu, Assistant Professor, Cornell University, 223 Rhodes Hall, Ithaca, NY, 14853, United States, [email protected] 1 - Dynamic Empty Car Management in Railyards Tolga Bektas, Postdoctoral Research Fellow, Universite de Montreal, Centre de Recherche sur les Transports, CP 6128, Succursale Centre-Ville, Montreal, Quebec, H3C 3J7, Canada, [email protected], Teodor Gabriel Crainic
This paper is concerned with accelerating the flow of empty cars in a major classification railyard in North America. Due to fluctations in inbound traffic, empty cars experience long delays at the yard. Such delays can be prevented by dynamic car reassignments. We describe a procedure using a sliding time window in order to reduce the time that empty cars spend in the yard and present the associated computational results.

3 - Information Technology and Firm Efficiency: A Non-Parametric Analysis Indranil Bardhan, Assistant Professor, The University of Texas at Dallas, School of Management, SM 41, 2601 N. Floyd Road, Richardson, TX, 75083-0688, United States, [email protected], Shu Lin, Sunil Mithas
We apply data envelopment analyses to study the impact of information technology investments across a large sample of global firms from 1998 to 2004. Using panel data on IT spending, we explore the impact of IT on firm efficiency and study trends in firm efficiencies over time.

I SA38
RASIG Student Paper Session: Top 3
Sponsor: Railroad Applications Sponsored Session
Chair: Pooja Dewan, BNSF Railway, Fort Worth, TX, United States, [email protected] 1 - RASIG Student Paper Competition
RASIG (Rail Applications Special Interest Section), a subdivision of INFORMS (Institute for Operations Research and Management Science) sponsored a student research paper contest on Management Science in Railroad Applications. Authors of First Place, Second Place, and Honorable Mention papers will be asked to present at this session.

2 - Dynamic Simulation Assignment Model for Intermodal MultiProduct Freight Transportation Hani Mahmassani, University of Maryland, 3128 J. H. Kim Building, College Park, MD, 20742, United States, [email protected], Kuilin Zhang, Jing Dong, Chung-Cheng Lu
A simulation-assignment model with joint mode and route choice is developed for multi-product freight shipments. Terminal operations are modeled as bulk queueing systems. Intermodal route choice is addressed by solving a multiproduct intermodal shortest path problem. The model is applied to a transEuropean corridor linking the Nordic countries to Southeastern Europe.

2 - First Place Winner: Spatial Scheduling and Resource Selection Problem: Modeling, Algorithm, and Application in the Production Gang Scheduling for Railway Maintenance Operations Gang Li, PhD Student, The University of Texas at Austin, McCombs School of Business, 1 University Station B6000, Austin TX 78712, United States, [email protected]
Given a set of geographically-dispersed jobs with time windows and inter-job coordination requirements, we address the problem of assigning and routing resources to jobs and sequencing the jobs to minimize the total cost for deploying and re-positioning resources. We formulate this combined routing-scheduling problem as a large-scale integer program, develop valid inequalities to strengthen the model, and report its successful application in the production gang schedul-

62

INFORMS PITTSBURGH— 2006
ing problem for a major railway company in North America. The model has been helping the company to save millions of dollars each year and significantly improve the quality of the company’s maintenance operations.

SA41

3 - Second Place: Modeling Reordering and Local Rerouting Strategies to Solve Train Conflicts During Rail Operations Andrea D’Ariano, Delft University of Technology, Faculty of Civil Engineering and Geosciences Transport and Planning Department, Stevinweg 1, 2628 CN Delft, Netherlands, [email protected]
A decision support system is presented to manage real-time timetable perturbations and blocked tracks. A mathematical model is adopted to solve train conflicts through reordering and local rerouting (dynamic use of platform tracks and alternative corridors). Rolling stock and passenger connections are also formulated. An iterative reordering and rerouting algorithm is developed to increase the punctuality. Experiments on a Dutch railway network show that the algorithm improves the solutions provided by practical dispatching rules.

3 - Service Disciplines in Location Set Covering Problems with Stochastic Demands and Congestion Seokjin Kim, Assistant Professor, Department of Business Administration, Millersville University, PO Box 1002, Millersville, PA, 17551-0302, United States, [email protected], Oded Berman, Dmitry Krass
We consider the problem of minimizing the number of servers to be located on a network and achieving the service level at each node at least a required level. Markovian demands originate and infinite capacity queues are positioned at each node. Servers can be stationed at nodes and serve calls within a pre-specified service radius. We investigate the effect of dynamic and static disciplines, and further develop location models for emergency medical services with two call classes.

4 - Honorable Mention: Solving Real Life Locomotive and Fuel Service Problems Balachandran Vaidyanathan, University of Florida, 303 Weil Hall, Gainesville FL 32611, United States, [email protected]
We study the problem of generating fuel and service feasible routing of locomotives starting from a base fleet assignment plan which satisfies several operational and contractual constraints. Fueling feasibility requires that every locomotive visits a service location at least once for every 1800 miles of travel. A locomotive fleet assignment plan which does not take the fueling and servicing requirements into consideration cannot be implemented and has limited value. Solving this problem is therefore of great value to US Railroads.

4 - The Probabilistic 1-Maximal Covering Problem on a Network with Discrete Demand Weights Jiamin Wang, Assistant Professor, Long Island University, Roth Hall 202, 720 Northern Boulevard, Brookville, NY, 11548, United States, [email protected], Oded Berman
We investigate locating a facility on a network with uncertain demand generated from the nodes. It is assumed that demand weights are independent discrete random variables. The objective is to maximize the probability that the total demand within a given distance of the facility reaches at least a pre-selected threshold level. It is shown that the problem is NP-hard and that an optimal solution exists in a finite set of dominant points. Solution procedures are developed to solve the problem.

I SA39
Where Do We Want to Go in Practice?
Cluster: Where Do We Want to Go?
(In observance of Arthur Geoffrion’s retirement)

I SA41
Queueing Methods and Applications
Sponsor: Applied Probability Sponsored Session
Chair: Victor Araman, Assistant Professor, New York University, Henry Kaufman Management Center, 44 West 4th Street, New York, NY, 10012, United States, [email protected] 1 - SRPT Scheduling and Large Deviations Bert Zwart, Associate Professor, H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology, 765 Ferst Drive, Atlanta, GA, 30332, United States, [email protected]
In recent years, the SRPT discipline has become highly relevant as potential scheduling discipline in, for example, web servers. Motivated by this, we consider the GI/GI/1 queue, and investigate the probability that a large sojourn time occurs under the SRPT discipline. We show that SRPT outperforms FIFO for heavy-tailed service times and that the reverse is the case for light-tailed service times. The results provide qualitative insights into the way large sojourn times occur.

Invited Session
Chair: Dwight Collins, [email protected] Co-Chair: Roundtable Strategic, Planning Committee, INFORMS, 7240 Parkway Drive, Suite 310, Hanover, MD, 21076-131, United States, [email protected] 1 - Where Do We Want to Go in Practice? Roundtable Strategic, Planning Committee, INFORMS, 7240 Parkway Drive, Suite 310, Hanover, MD, 21076-131, United States, [email protected]
The evolving global economy, developments in technology, and changing managerial practices provide the OR/MS profession with new opportunities and challenges. OR/MS must change along with the world. In this session, the Roundtable examines pertinent developments, identifies and prioritizes candidate strategic goals for OR/MS professionals, and presents its plans and suggestions for making progress in those directions.

I SA40
Location Analysis: Models, Algorithms and Applications
Sponsor: Location Analysis Sponsored Session
Chair: Jiamin Wang Assistant Professor, Long Island University, Roth Hall 202, 720 Northern Boulevard, Brookville, NY, 11548, United States, [email protected] 1 - Integrated Location, Routing and Scheduling Problems: Models and Algorithms Zeliha Akca, PhD Candidate, Lehigh University, Industrial and Systems Engineering, Bethlehem, PA, 18015, United States, [email protected], Rosemary Berger
We study the problem of determining the locations of capacitated facilities within a distribution network; deliveries are made along multiple stop routes and multiple routes may be assigned to the vehicles, which have capacity and time restrictions. We develop a set-partitioning-based formulation and identify valid inequalities. We develop a branch-and-price algorithm to solve instances of the problem and present computational results.

2 - Dynamic Control of a Make-to-Order, Parallel-Server System with Cancellations Baris Ata, Assistant Professor, Kellogg School of Management, Northwestern University, 2001 Sheridan Road, Evanston, IL, 60208, United States, [email protected], Melanie Rubino
Motivated by make-to-order production systems, we consider a dynamic control problem for a multiclass, parallel-server queueing system. The production system serves multiple classes of customers who require rigid due-date lead times and may cancel their order subject to a cancellation penalty. Under heavy traffic conditions, an approximately optimal non-greedy outsourcing and resource allocation policy is derived.

3 - Importance Sampling for Credit Risk Models Achal Bassamboo, [email protected], Sachin Jain
In recent times, credit markets have grown and research has been focused on modeling of default events for multi-firms. Most of these models rely on simulation for the pricing of derivatives. We consider one such class of models; specifically the intensity driven doubly stochastic processes and propose an asymptotically optimal change of measure for efficient simulation. We apply our result to specific models and support our theory with numerical simulations.

2 - Locating Cell Phone Towers in a Rural Environment H. A. Eiselt, University of New Brunswick, Fredricton, NB, Canada, [email protected], Vladimir Marianov
The presentation investigates the location of cell phone towers in a district in Nepal. In addition to provide widespread coverage, planners must also guard against losses due to insurgents. Our objective is a combination of average survival and worst-case survival of towers in case one of the towers is destroyed.

4 - Stein’s Method for Some Stochastic Equations Jose Blanchet, Assistant Professor, Harvard University, Statistics Department, 1 Oxford Street, Cambridge, MA, 02138, United States, [email protected]
In 1972, Stein introduced a powerful method for approximating expectations via two ingredients: 1) An operator that characterizes the approximating distribution and 2) A coupling among random elements of interest. We will explain Stein’s method and develop approximations for solutions of some important stochastic equations. Our examples include ARCH models (widely used in finance) and geometric sums, which arise in the study of the M/G/1 queue, the classical risk model and reliability models.

63

SA42 I SA42
Logistics Scheduling
Cluster: Scheduling Invited Session

INFORMS PITTSBURGH — 2006
3 - When Should the Majority of Remaining US Combat Troops Withdraw from Iraq? Tori Hinde, KGSB/University of Pittsburgh, Mervis Hall, Pittsburgh, PA, United States, [email protected]
The goal of this project was to determine the optimal US troop withdrawal strategy. The synthesized results of the entire ANP model which took into account the weighting factors of the strategic rating model indicate that the highest rated alternative for the short run is to withdraw major US combat troops in 2-5 years. The long run results, and those that should be used, indicate that the alternative with the best results for all participants is to redeploy the US troops within the region.

Chair: Chung-Yee Lee, Head & Chair Professor, Hong Kong University of Science & Technology, Clear Water Bay Road, Kowloon, Hong Kong, China, [email protected] 1 - A Nash Bargaining Solution for Partition of Jobs Between Two Manufacturers Xiaoqiang Cai, Professor, Department of Systems Engineering and Engineering Management, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong, [email protected], Quanle Chen
Two manufacturers negotiate to partition a set of jobs that they have jointly obtained. Each manufacturer has his own processing capacity and setting, and thus has to look for the best schedule for his processing given any subset of jobs that he may get. We develop an algorithm, and show that it can derive the best solution to partition the jobs in a mutually acceptable and beneficial manner.

4 - A Complex Process of Forming a Regional Financial Arrangement (RFA) Iwan Azis, Professor, Johnson Graduate School of Management, Cornell University, 215 W. Sibley, Ithaca, NY, United States, [email protected]
In East Asia, the most decisive event that created interest for a closer regional cooperation was the 1997/98 financial crisis. Ineffective policy responses to a crisis, slow progress of WTO, growing risks of global imbalance, and strong interest among business sectors to access foreign markets and capital, are among the important reasons for such proliferations. The preferred form of cooperation developed using the ANP suggests that RFA w/o a common currency basket is most suitable.

2 - Non-Cooperative Outsourcing Games Tolga Aydinliyim, Case Western Reserve University, Weatherhead School of Management, 11119 Bellflower Road, Cleveland, OH, 44106, United States, [email protected], George Vairaktarakis
We present a non-cooperative outsourcing model where each manufacturer processes part of his jobs on a dedicated in-house machine and sends his remaining jobs to a third-party. The third-party has a flexible machine which can process jobs from all manufacturers. Each manufacturer’s strategy is determined as the set of jobs to be processed on the flexible machine. The Nash Equilibrium schedule and the centralized schedule do not necessarily coincide, so savings due to coordination are possible. We present structural properties, algorithms and coordination results.

I SA44
Applied Probability I
Contributed Session
Chair: Ivan Guardiola, Department of Industrial Engineering, Texas Tech University, Box 43061, Lubbock, TX, 79409, United States, [email protected] 1 - Stationary Work-Load of G/M/1 Queueing System with a Hysterestic Service Policy Sunggon Kim, Department of Information and Statistics, Gyeongsang National University, Jinju, South Korea, [email protected], Jongho Bae
We consider a queueing system with general inter-arrival times and exponentally distributed service times. The service policy has two stage. Server starts to serve the customer with normal service rate of 1. As soon as the amount of total workload is larger than a threshold, the server increases the rate to a high rate of M, and keeps this rate until there are no customers. The stationary work-load is obtained by using level-crossing argument, martingale approach, and decomposition technique.

3 - Single Machine Scheduling Under Potential Disruption Chung-Yee Lee, Head & Chair Professor, Hong Kong University of Science & Technology, Clear Water Bay Road, Kowloon, Hong Kong, China, [email protected], Gang Yu
In this paper, we study a new scheduling problem in which there is a possibility that a disruption will occur at a particular time and if it happens the machine (facility) will become unavailable for certain duration with a certain probability. We study two objective functions: the expected total weighted completion times and the expected maximum tardiness. (This research is supported in part by Hong Kong RGC grant HKUST 6145/03E.)

4 - Competitive Agent Scheduling and Its Applications Michael Pinedo, New York University, IOMS Department, 44 West 4th Street, New York, NY, 10012, United States, [email protected], Joseph Y. Leung
Two agents have to schedule two sets of jobs on a single resource. The two agents have different constraints and objectives. We consider the problem of minimizing the value of the objective function of agent A, while keeping the value of the objective of agent B less than or equal to a fixed value Q.

2 - The Use of Non-Stationary Palm Measures in Queueing Theory Brian Fralix, PhD Student, ISyE, Georgia Institute of Technology, 324592 Georgia Tech Station, Atlanta, GA, 30332, United States, [email protected], Richard Serfozo, German Riano
We show that, under non-stationary Palm measures, some classical results from queueing theory are in a sense true outside of a stationary framework (in particular, Little’s law and Papangelou’s lemma (which gives us a time-dependent version of PASTA)). Moreover, we show that under certain conditions, these Palm probabilities of events converge in some sense to their stationary counterpart. In particular, when analyzing Markov processes, these probabilities often exhibit a nice form.

I SA43
What Would Not Have Been Possible Without the Work of Thomas L. Saaty
Cluster: Analytic Hierarchy Process Invited Session
Chair: Iwan Azi, Professor, Johnson Graduate School of Management, Cornell University, 215 W. Sibley, Ithaca, NY, United States, [email protected] 1 - On Validating the Analytic Hierarchy/Network Processes Rozann Saaty, Creative Decisions Foundation, Pittsburgh, PA, United States, [email protected]
An essential fact regarding the truth of a theory is its validation in practice. One validates with: (1) A pairwise comparison matrix, (2) A hierarchy, (3) A network (both simple with market share and with benefits, opportunities, costs and risks. A compatibility index is used to compare the derived results with data.

3 - On the Stability of Moment Closure Methods under Various Closure Schemes Ivan Guardiola, Department of Industrial Engineering, Texas Tech University, Box 43061, Lubbock, TX, 79409, United States, [email protected], Dr. Timothy I Matis
In this presentation, we highlight current research findings related to the stability of moment closure methods under various closure schemes. We propose a basic stability checking methods, as well as an interpretation of the stability of a closure scheme based on the kurtosis.

4 - Analysis of Unreliable Finite Buffer Multi-Server System with Vacations Zhe George Zhang, Professor, Western Washington University, 516 High Street, Bellingham, WA, 98225, United States, [email protected], Jau-Chuan Ke
We consider a finite M/M/c queueing system in which all servers are unreliable and operate (d, c) vacation policy. The steady-state probability distribution of system size is calculated by means of the QR factorization with Cramer’s rule. A computational algorithm is developed to determine the optimal vacation policy (d, c) to maximize the profit under a cost and revenue structure.

2 - Advanced Decision Technologies Daniel Saaty, President, Decision Lens, Inc., 282 N. Washington Street, Falls Church, VA, 22046, United States, [email protected]
Although AHP/ANP have been in existence for many years, there continues to be a lack of simple and elegant decision support software applications available to business users. Here we give a presentation on the latest in Decision Support software applications employing AHP/ANP and Linear Optimization to support government and commercial trade-off and resource allocation decisions. The software is shown in real business example for resource allocation in group decision making at DOD.

64

INFORMS PITTSBURGH— 2006
5 - Improved Inspections for Systems with Non-Self-Announcing Failures Bahadir Aral, PhD Candidate, Texas A&M University, 241 Zachry 3131 TAMU, College Station, TX, 77843, United States, [email protected], Georgia-Ann Klutke
In this study, we consider detection strategies for systems with non-selfannouncing failures. System failures occur according to a non-homogeneous Poison Process with a known intensity function. We express the limiting average availability and the limiting inspection rate per unit time for periodic inspections and we also show how to construct an better inspection strategy, which will have a higher limiting availability, without changing the inspection rate.

SA48

study two models that consider each of these fixed costs. For both models, we construct a lower bound to the optimal cost, an optimization algorithm, and an effective heuristic.

3 - Inventory Planning for Low-Demand Items for a Multi-Warehouse Online Retailer Ping Josephine Xu, Operations Research Scientist, Amazon.com, 605 5th Avenue South, Seattle, WA, 98104, United States, [email protected], Stephen Graves, Russell Allgor
E-tailers stock many low-demand SKUs in multiple warehouses to offer an extremely wide range of selections with very fast customer service at a minimum cost. The e-tailer manages and controls the inventory centrally: it utilizes its entire inventory, regardless of location, to serve demand. For a given stocking plan, we model the expected transportation and inventory costs to serve demand. We develop guidelines for stocking low-demand items across multiple warehouses.

I SA45
Tutorial: Location, Transport and Land-Use: Modeling Spatial-Temporal Information
Cluster: Tutorials Invited Session
1 - Location, Transport and Land-Use: Modeling Spatial-Temporal Information Yupo Chan, University of Arkansas at Little Rock, 2801 South University Avenue, Department of System Engineering, Little Rock, AR, 72204-1099, United States, [email protected]
This tutorial identifies principles that govern siting, community development, and product/service delivery. Included are procedures to perform: site location, land-use planning, location-routing, competitive allocation of products & services, and spatial forecasting. We develop solution techniques for a number of applications, including emergency-response, supply-chain management, environmental planning, intelligent transportation systems and supply-chain management.

4 - Analytical Insights into the Two-Stage Supply Chain Configuration Problem Sean Willems, Assistant Professor of Operations Management, Boston University, School of Management, SMG 673, 595 Commonwealth Avenue, Boston, MA, 02215, United States, [email protected], N. Grace Hua
This research demonstrates how leadtime and cost impact total supply chain landed cost, inventory placement and the optimal supply chain configuration. We prove the relative importance of cost of good sold versus inventory cost drives whether or not a high cost low leadtime option is preferred; also, we show that there is benefit to synchronize the supply chain. Finally, we analytically prove the downstream stage is more likely to choose a high cost low leadtime option than the upstream stage.

I SA48
Retail Supply Chain Management
Sponsor: Manufacturing & Service Operations Management Sponsored Session
Chair: Naren Agrawal, Associate Professor, Santa Clara University, OMIS Department, Santa Clara, CA, 95070, United States, [email protected] Co-Chair: Stephen Smith, J. C. Penney Professor, Santa Clara University, OMIS Department, 500 El Camino Real, Santa Clara, CA, 95053-0382, United States, [email protected] 1 - Inventory Management of a Fast-Fashion Retail Network Jeremie Gallien, Associate Professor, Sloan School of Management, Massachusetts Institute of Technology, E53-389, 30 Wadsworth Street, Cambridge, MA, 02142, United States, [email protected], Felipe Caro
Fast-fashion retailers like Zara have met some success responding to volatile demand trends through frequent introductions of new garments produced in small series. An important associated operational problem is the allocation over time of a limited amount of inventory across all stores in their network. We present stochastic and deterministic models developed in collaboration with a large fast-fashion retailer to address this challenge, then discuss the implementation and impact of this work.

I SA46
Tutorial: Pricing and Revenue Management
Cluster: Tutorials Invited Session
1 - Pricing and Revenue Management Andrew Boyd, Chief Scientist and Senior Vice President Science and Research PROS, 3100 Main Street, Suite 900, Houston, TX, 77002, United States, [email protected]
Pricing and revenue management has become one of the fastest growing specialties within OR. Businesses are investigating what mathematical modeling can bring to the world of sales and distribution.In this tutorial, we introduce some basic pricing and revenue management models and their interactions. We also examine the commercial market for pricing and revenue management software and services and identify future research opportunities.

I SA47
Inventory and Logistics Management
Cluster: Supply Chain and Operations Engineering Invited Session
Chair: Brian Tomlin, Assistant Professor, University of North Carolina, McColl Building, CB3490, Chapel Hill, NC, 27599, United States, [email protected] 1 - On the Value of Pre-Season Orders when Supply is Uncertain Yimin Wang, University of North Carolina, Chapel Hill, KenanFlagler Business School, McColl Building, CB 3490, Chapel Hill, NC, 27599, United States, [email protected], Brian Tomlin
We study the ordering decision of a firm that faces combined supply and demand risks. The supply risk refers to long and unreliable procurement leadtimes and the demand risk refers to the firm’s uncertainty about its pre-season and inseason orders. We characterize the firm’s optimal timing and quantity decision. We prove that a firm with pre-season orders might order earlier than one without. We also prove that a firm may strictly prefer a less reliable to a more reliable leadtime.

2 - Supplier Competition in Retail Supply Chains Gurhan Kok, Duke University, One Towerview Drive, Durham, NC, United States, [email protected]
A retailer determines the stocking quantities of several competing products that are provided by different suppliers. Each supplier can offer better contract terms to the retailer in order to increase the stocking quantity of her product. We study the competition between the suppliers under different types of contracts and report its impact on contractual terms and the profits of each party involved.

3 - Optimal Retail Assortments for the Nested MNL and Other Consumer Choice Models Stephen Smith, J. C. Penney Professor, Santa Clara University, OMIS Department, 500 El Camino Real, Santa Clara, CA, 950530382, United States, [email protected]
This paper develops methods for determining optimal retail assortments, given that heterogeneous customers choose products based on the nested multinomial logit model. Optimal assortments for consumer choice models based on “satisficing” are also derived. A commercial database of consumer preferences is used to illustrate the optimization methods and to compare the corresponding assortments.

2 - Analysis of Serial Inventory Systems with Batch Ordering and Nested Replenishment Schedule Kevin Shang, Fuqua School of Business, Duke University, Durham, NC, United States, [email protected], Sean Zhou
We consider an N-stage serial inventory system where material is ordered in batches and according to a nested schedule. There are two types of fixed cost considered in this system: setup cost and order cost. The former is associated with each batch ordered; the latter is incurred for each replenishment cycle. We

65

SA49 I SA49
History of OR/MS I

INFORMS PITTSBURGH — 2006 I SA51
Data Envelopment Analysis I
Contributed Session
Chair: Olga Kaminer, PhD Student, York University, 4700 Keele Street, Toronto, ON, M3J 1P3, Canada, [email protected] 1 - An Industry-Level Analysis of the Potential and Realized Value of IT Kim Huat Goh, University of Minnesota, 321 19th Avenue South, Minneapolis, MN, 55455, United States, [email protected], Robert Kauffman
IT investments are evaluated by weighing their potential benefits and actual costs. Potential benefits, however, do not always translate to actual payoffs. We adopt the Malmquist productivity index to model the potential and realized value of IT and apply data envelopment analysis techniques to solve the model. We examine the gap between potential and realized value of IT in industries within the United States and assess the impact of competition on the potential and realized value of IT.

Cluster: History of OR and Management Science Invited Session
Chair: Maurice Kirby, Lancaster University, Lancaster, United Kingdom, [email protected] 1 - Leonid Khachiyan (1952-2005) Endre Boros, Professor, RUTCOR, Rutgers University, 640 Bartholomew Road, Piscataway, NJ, 08854-8003, United States, [email protected]
Leo Khachiyan became famous for proving first the polynomiality of linear programming. He was in fact a genius mathematician and computer scientist who also made major contributions to several other areas, including game theory, combinatorial optimization and the theory of data mining. In this short talk we focus on his contributions in the last decade.

2 - Management and Probability Theory at the Bell System: 1900-1929: Traffic Operations and Quality Control Paul J. Miranti, Professor, Rutgers Business School, 111 Washington Street, Newark, NJ, 07102, United States, [email protected]
This paper analyzes two antecedents to the emergence of the operations research and management science disciplines that evolved from the adaptation of probability theory to resolve problems in traffic management and manufacturing quality control at the Bell System prior to 1929.

2 - Benchmarking the Competitiveness of Nations Through Neural Networks and Data Envelopment Analysis Özgür Kabak, Istanbul Technical University, Isletme Fakultesi, Macka, Istanbul, Turkey, [email protected], Emel Aktas, Füsun Ülengin, Gündüz Ulusoy, Sule Önsel
Most composite indicators currently used to compare the competitiveness of nations either take predetermined fixed weight values or use subjective weights for different cluster of nations. In this study, the criteria that play the dominant role in each cluster are revealed based on the weights that’ve been specified by using artificial neural networks. A new composite index is proposed for the ranking of the nations. Data Envelopment Analysis is used to analyze the relative position of the nations.

3 - Operations Research and the Area Bombing of Nazi Germany Maurice Kirby, Lancaster University, Lancaster, United Kingdom, [email protected]
This paper revisits the author’s original research on the role of Operations Research (OR) in informing the tactics and strategy of the Allied airforces in relation to the campaign of “area attack” launched against Germany in 1942. Contributions to the literature on the “morality” of area bombing and its contribution to Germany’s defeat are ongoing, but the role of OR in justifying the bombing of civilian targets is rarely acknowledged. This paper reasserts the case for area bombing as a moral and entirely rational means of engaging with Nazi Germany in advance of the D-Day landings.

3 - Performance Evaluation and Market Potential Forecasting Based on Statistical and DEA Methods Wenjun Yin, IBM China Research Laboratory, Building 19 Zhongguancun Software Park, Beijing, BJ, China, [email protected], Jin Dong
Banking branches performance evaluation and market potential forecasting plays key roles in branch net-works optimization. However, current works focus only on internal business of the branches and ignore the geo-graphic and demographic information around the branches. With linkage to the geographic information system (GIS), statistical tools, data envelopment analysis (DEA) and their crossover are applied. Good result was achieved by a real case analysis of one of the biggest China banks.

I SA50
INFORMS Optimization Society Award Session
Sponsor: Optimization Sponsored Session
Chair: Tamás Terlaky, Professor, McMaster University, 1280 Main Street West, Hamilton, ON, L8S4K1, Canada, [email protected] 1 - The Farkas Prize of the INFORMS Optimization Society
The Farkas Prize of the INFORMS Optimization Society was established in 2006, and is administered by the Optimization Society (OS) of INFORMS. The Farkas Prize is awarded for the most significant contribution to the field of optimization by a researcher, or a team of researchers. It is awarded bi-annually at the INFORMS National Meeting.

4 - The Validity of Input Aggregation in DEA Models: A Nonparametric Statistical Test Kankana Mukherjee, Assistant Professor, Worcester Polytechnic Institute, Department of Management, 100 Institute Road, Worcester, MA, 01609, United States, [email protected], Subhash Ray
In many DEA applications multiple inputs or outputs are aggregated in value terms to reduce the number of constraints, thereby increasing the “degrees of freedom”. While the theoretical implications of such aggregation have been studied, the statistical side still is in a rudimentary stage. We propose a nonparametric statistical test of the validity of price-weighted input aggregation using a smoothed bootstrap procedure. This is illustrated with an application to Indian manufacturing data.

2 - The Optimization Prize for Young Researchers
The Optimization Prize for Young Researchers, established in 1998 and administered by the Optimization Society of INFORMS. It is awarded in even numbered years to one (or more) young researchers for the most outstanding paper in optimization that is submitted to or published in a refereed professional journal. The prize serves as an esteemed recognition of promising colleagues who are at the beginning of their academic or industrial career.

3 - The Legacy of Gyula Farkas András Prékopa, Professor, RUTCOR, Rutgers University, 640 Bartholomew Road, Piscataway, NJ, 08854, United States, [email protected]
Farkas’ Theorem and Lemma is probably the most frequently cited result in the optimization literature, it is one of the first results taught in introductory LP and OR courses, both on undergraduate and graduate levels. He was a forerunner of optimization theory. He formulated the necessary condition of optimality, in an analytical mechanical framework, proved his Lemma and obtained other fundamental scientific results. His works represent the importance of applications as well. This talk highlights Gyula Farkas’s life and work.

5 - Effect of Competitive Environment on the Choice of Manufacturing Capabilities Olga Kaminer, PhD Student, York University, 4700 Keele Street, Toronto, ON, M3J 1P3, Canada, [email protected], Markus Biehl, Wade Cook, Murat Kristal, Aleda Roth
There is an ongoing debate in the strategy literature about whether or not tradeoffs need to be made in the development of competitive capabilities. In this study we investigate the trade-offs debate by using data from 150 manufacturers across North America. We theorize under which environmental conditions, such as environmental dynamism, the trade-off and cumulative models hold. We test our hypothesis using a DEA approach of manufacturing firms’ capabilities and financial performance measures.

4 - The 2006 Winner of the Farkas Prize
The winner of the 2006 Farkas Prize will present a survey of his most influential results and give an insight of his recent research interest.

5 - The Winner of the 2006 Optimization Prize for Young Researchers
The winner of the 2006 Optimization Prize for Young Researchers presents the award winning paper.

66

INFORMS PITTSBURGH— 2006 I SA52
Customer Behavior in Revenue Management
Sponsor: Revenue Management & Pricing Sponsored Session
Chair: Xuanming Su, Assistant Professor, University of California, Haas School of Business, Berkeley, CA, 94720, United States, [email protected] 1 - Dynamic Pricing of Limited Inventories when Customers Negotiate Chia-Wei Kuo, Department of Industrial and Operations Engineering, University of Michigan, Ann Arbor, MI, 48109, United States, [email protected], Goker Aydin, Hyun-Soo Ahn
We consider a firm who adjusts the sticker price dynamically in the course of a selling season. Each customer observes the posted price and engages in a negotiation process to complete the purchase. We explore the interaction between dynamic pricing and negotiation and show how inventory and remaining selling period affect the seller’s optimal pricing policy.

SA54

2 - Learning Customers’ Preferences through Repetitive Sale Interactions Paulo Rocha e Oliveira, Assistant Professor of Marketing, University of Miami, United States, [email protected], Rene Caldentey, Gabriel Bitran
Knowing customers’ preferences enhances a firm’s ability to design and offer the right menu of products. Learning is not free, and companies must allocate valuable resources if they really want to know their customers better. In this paper, we investigate this trade-off between the value and cost of learning. We use an axiomatic approach to develop a learning model and investigate how pricing policies can be used to speed up the learning process while maximizing the firm’s financial performance.

3 - Dynamic and Data-Driven Inventory Management Under Uncertainty Aurelie Thiele, P.C. Rossin Assistant Professor, Lehigh University, 200 W. Packer Avenue, Bethlehem, PA, 18015, United States, [email protected]
We propose a framework based on dynamic and data-driven optimization to incorporate demand uncertainty. The approach is based on the monitoring of the actual demand and the decision-maker’s ability to address sudden changes quickly, for instance by increasing the frequency of the observations and tracking additional quantities such as demand for other products or at other stores. We provide theoretical insights on the optimal solution and illustrate our findings on a numerical example.

2 - The Effects of Information on a Queue with Balking and Phase-Type Service Times Pengfei Guo, Ph.D. Candidate, Duke University, 1 Towerview Drive, Durham, NC, 27708, United States, [email protected], Paul Zipkin
In this paper, we generalize the models in Guo and Zipkin (2003) to a system with phase-type service times. We model the system with no, partial (the system occupancy) and full (the exact waiting time) information about delay. We compute the busy probability and customers’ average utility numerically. Our numerical results show that whether information is good or bad for the server or customers depends on the shape of customers’ distribution.

4 - Centralized vs. Decentralized Competition for Price and Lead-time Sensitive Demand Pelin Pekgun, PhD Candidate, Georgia Institute of Technology, 765 Ferst Drive, Atlanta, GA, 30332, United States, [email protected], Pinar Keskinocak, Paul Griffin
We study two firms that compete on the basis of their price and leadtime decisions in a common market. We explore the impact of the decentralization of price and leadtime decisions, as quoted by the marketing and production departments, respectively, comparing three scenarios: (i) Both firms are centralized, (ii) Only one firm is centralized, (iii) Both firms are decentralized.

3 - On the Threshold Purchasing Behavior of Customers Facing Dynamically Priced Perishable Products Yong-Pin Zhou, University of Washington, Business School, Seattle, WA, United States, [email protected], Ming Fan, Minho Cho
Dynamic pricing is a standard practice that firms use for revenue management. As price fluctuates with the changes in inventory and time, a strategic customer may postpone a purchase in anticipation of lower prices in the future. We analyze a threshold purchasing policy for the strategic customer, and use numerical studies to study its impact on both the strategic customer’s utility and the firm’s revenue. We find that in most cases the policy can benefit both.

I SA54
Interfaces Between Marketing, Operations and Technology I
Cluster: Operations and Marketing for Emerging Markets Invited Session
Chair: Jiong Sun, Doctoral Student, Tepper School of Business, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected] Co-Chair: Sunder Kekre, Professor of Operations Management and Manufacturing, Tepper School of Business, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected] 1 - Product Line Design: Variety, Mass Customization and Economies of Scale Dennis Z. Yu, Assistant Professor, Clarkson University, School of Business, PO Box 5790, Potsdam, NY, 13699-5790, United States, [email protected], Sergio Chayet, Panos Kouvelis
A manufacturer’s product line design is studied when the production technology allows quality differentiation. Customers are heterogeneous in their valuation of product quality. The manufacturer faces a tradeoff between production capacity investment and holding cost of WIP due to the system congestion. The presence of economies of scale is characterized by the existence of the product variety limit. We derive the manufacturer’s optimal strategies under both dedicated and flexible technologies.

4 - Waiting Till the Last Minute: Strategic or Irrational? Xuanming Su, Assistant Professor, University of California, Haas School of Business, Berkeley, CA, 94720, United States, [email protected]
Consumers often wait until the last minute before making a purchase. This indicates strategic behavior if they can secure a last-minute discount. Yet, consumers sometimes wait even when this implies higher prices and lower availability. We adopt a behavioral approach towards this phenomenon and study its implications on dynamic pricing strategies.

I SA53
Revenue Management and Beyond
Sponsor: Revenue Management & Pricing Sponsored Session
Chair: Felipe Caro, Assistant Professor, UCLA Anderson School of Management, 110 Westwood Plaza, Suite B420, Los Angeles, CA, 90095, United States, [email protected] Co-Chair: Rene Caldentey, Assistant Professor in Operations Management, Stern School of Business, New York University, 44 West Fourth Street, KMC 8-77, New York, NY, 10012, United States, [email protected] 1 - Integer Programming Models of Pricing Problems with Probabilisitic Choice Romy Shioda, University of Waterloo, 200 University Avenue, Waterloo, ON, Canada, [email protected], Tuncel Levent
We present mixed-integer programming formulations of several product pricing models using probabilistic customer choice models. In these models, the probability that a customer segment would purchase a given product depends on their reservation prices and the product prices. We present some mathematical properties of these problems and computational experiments.

2 - Fostering New Technology Adoption Ozgen Karaer, PhD Candidate, Management Science and Engineering Department, Stanford University, Stanford, CA, United States, [email protected], Feryal Erhun, Hau L. Lee
Faced with a new technology introduced in a market, a firm considers uncertainties regarding the future of this technology, network effects of adoption, and diseconomies of scales in costs while making the adoption decision. Recently, adoption of RFID technology is lower than expected due to this individual perspective of each firm. We study this low adoption problem and analyze the impact of methods to foster adoption such as quantity discounts, external funds, priority schemes, and mandates.

67

SA55

INFORMS PITTSBURGH — 2006
Supply and Advance Demand Information Mehmet Sekip Altug, PhD Candidate, Columbia University, Graduate School of Business, Uris Hall 4J, New York, NY, 10027, United States, [email protected], Alp Muharremoglu
In this paper, we combine both sides of information sharing and consider a manufacturer who gets both “advance supply” and “advance demand” information. The main questions we study are: i) How does the manufacturer use this information? ii) What is the value of such information sharing? We show that state-dependent base stock policies are optimal for the manufacturer. We derive closed-form approximations and analyze the value of collaboration.”

3 - Imitative Competition: Implications for Strategic Sourcing and Technology Transfer Jiong Sun, Doctoral Student, Tepper School of Business, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected], Laurens Debo, Sunder Kekre, Jinhong Xie
Technology transfer offers global firms an opportunity to penetrate into emerging markets as well as source for their home markets. However, it also poses risks of potential technology imitation by local firms who may enter the emerging market and invade subsequently the global firms’ home markets as well. We study how the competition affects a global firm’s technology transfer and sourcing decisions, and a local competitor’s imitation and exporting choices.

4 - Products with Network Externality: The Relevance of Unbundling and Mixed Bundling R. Venkatesh, Associate Professor, University of Pittsburgh, Katz Graduate School of Business, 332 Mervis Hall, Pittsburgh, PA, 15260, United States, [email protected], Vijay Mahajan, Ashutosh Prasad
Despite the hype surrounding Microsoft’s decision to bundle Internet Explorer and Windows, we argue that unbundling and mixed bundling can be optimal for products with network externality. Specifically, we analyze a stylized model to propose normative guidelines on optimal bundling strategies for a profit maximizing seller. We show that when products have low to moderate network externality, unbundling is better when marginal costs are high. Mixed bundling is also superior to pure bundling except when the products have strong network externality.

Sunday, 10:00am - 11:30am
I SB01
Applications of the Arora-Rao-Vazirani Algorithm
Sponsor: Optimization/ Network and Combinatorial Optimization Sponsored Session
Chair: Assaf Naor, Microsoft Research, One Microsoft Way, Redmond, WA, 98052-6399, United States, [email protected] 1 - Geometric Embeddings and Sparsest Cut Variants James R. Lee, Professor, Institute for Advanced Study, 1 Einstein Drive, Princeton, NJ, 08540, United States, [email protected]
In general, problems in the Sparsest Cut genre involve finding balanced separators in graphs: Our task is to break a given graph into two “large” pieces while minimizing the “interface” between them. Depending on the precise definition of these two objectives, different types of problems arise. We will discuss these variants and their applications, along with the known algorithmic results, and the relationship to semi-definite optimization and geometric embeddings.

I SA55
Supply Chain Logistics I
Contributed Session
Chair: Mehmet Sekip Altug, PhD Candidate, Columbia University, Graduate School of Business, Uris Hall 4J, New York, NY, 10027, United States, [email protected] 1 - Supply Chain Design on a Line with Lateral Transshipments Yale Herer, Professor, Technion, Industrial Engineering and Management, Haifa, 32000, Israel, [email protected], Hussein Naseraldin
We investigate the problem of simultaneously determining the number of retail outlets, their location, the allocation of customers to retail outlets, inventory policy, transshipment policy, inventory quantities, and transshipment quantities in a periodic infinite horizon stochastic inventory problem where the objective is to minimize the average cost per period. We obtain the optimal solution and discuss its properties.

2 - New Approximation Guarantee for Chromatic Number Eden Chlamtac, Research Assistant, Princeton University, Computer Science Building, 35 Olden Street, Princeton, NJ, United States, [email protected], Moses Charikar, Sanjeev Arora
We describe an algorithm which colors any 3-colorable graph using O(n0.2074) colors, thus improving the algorithms of Karger, Motwani and Sudan, and Blum and Karger, which used O(n3/14) colors. Using ideas inspired by the work of Arora, Rao and Vazirani, we demonstrate a better relation between vector chromatic number and true chromatic number. While this gives some improvement, we obtain our best result by adding “odd-cycle constraints” to the semidefinite program.

2 - Container Terminals as an Integral Part of Arab and Global Supply Chains: The Case of Lebanon Nouri Beyrouti, Assistant Professor in Marketing & Management, Lebanese American University, Chouran 1102 2801, Beirut, 135053, Lebanon, [email protected]
With ever increasing containerization, the number of seaport container terminals and the competition among them has become quite remarkable. Operations are nowadays unthinkable without effective and efficient use of information technology as well as appropriate optimization (operations research) methods. The research will describe the main logistics process and operations in container terminals and provide an understanding of Lebanon competitive advantage.

3 - Directed Metrics in Graph Optimization Problems Yury Makarychev, Princeton University, 35 Olden Street, Princeton, NJ, 08540, United States, [email protected], Moses Charikar, Konstantin Makarychev, Amit Agarwal
The theory of metric embeddings has provided a powerful toolkit for undirected graph partitioning problems. The metrics of interest arise naturally as solutions of mathematical programming relaxations for graph partitioning problems. No analog of this embedding theory exists for directed (asymmetric) metrics, the natural distance functions that arise in considering directed problems. We discuss some successes and some obstacles in extending the embedding machinery to directed metrics.

3 - A Single-Warehouse Multi-Retailer Inventory System with Discounts Based on Order Frequency Jose Ventura, Professor, Pennsylvania State University, 310 Leonhard Building, University Park, PA, 16802, United States, [email protected], Brian Rieksts, Yale Herer
We study a single-warehouse multi-retailer inventory system in which a product is bought by the warehouse from an external supplier, who gives all-unit discounts on the purchasing cost based on the average order frequency. From the warehouse, the product is dispersed to satisfy the demand of all the retailers. A model with assumptions analogous to the EOQ environment is developed and an algorithm for a power-of-two policy that is 94% effective for a variable base planning period is derived.

4 - A New Integrality Gap for the Semidefinite Relaxation of Sparsest Cut Assaf Naor, Microsoft Research, One Microsoft Way, Redmond, WA, 98052-6399, United States, [email protected]
In this talk I will show that the Heisenberg group has an equivalent metric which is in squared L_2, yet it does not embed into L_1. It follows that the classical Heisenberg geometry on R3 is a simple counter example to the Goemans-Linial conjecture on the integrality gap of the semidefinite relaxation of the Sparsest Cut Problem. Based on joint work with James R. Lee.

I SB02
Joint Session Open-Source/ICS: Applications of the Nonlinear Optimization Software IPOPT II
Cluster: Open-Source Software: Open Source, Open Standards, Open Data, INFORMS Computing Society Invited Session
Chair: Andreas Waechter, Research Staff Member, IBM TJ Watson Research Center, 1101 Kitchawan Road, Yorktown Heights, NY, 10598, United States, [email protected]

4 - Misplaced Inventory and RFID: Information and Coordination Almula Camdereli, UNC-Chapel Hill, 1800 Baity Hill Dr, Apt 215, Chapel Hill, NC, 27514, United States, [email protected], Jay Swaminathan
Misplaced inventory is a major operational problem in many supply chains. Adoption of this technology has a fixed cost and variable cost of implementation, which can cause incentive issues in the supply chain. In this paper, we consider centralized, uncoordinated and coordinated supply chains under misplacement of inventory subject to uncertain demand. We characterize the incentives of the parties in these supply chains to invest in RFID.

5 - Full Collaboration: Analyzing Inventory Systems with Advance

68

INFORMS PITTSBURGH— 2006
1 - Large-Scale Nonlinear Programming for Optimal Design and Operation of Simulated Moving Bed Processes Yoshiaki Kawajiri, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected], Lorenz Biegler
Simulated Moving Bed (SMB) chromatography is a widely used separation technique in chemical process industries, which is modeled by Partial Differential Algebra ic Equations (PDAEs). We present several optimization case studies of large-scale nonlinear programming of SMB with over 100,000 variables solved efficiently using IPOPT.

SB05

I SB04
Strategic Interactions in Supply Chains
Cluster: Supply Chain and Operations Engineering Invited Session
Chair: Haresh Gurnani, University of Miami, Department of Management, Coral Gables, FL, 33124, United States, [email protected] Co-Chair: Yunzeng Wang, University of Texas at Dallas, Richardson, TX, United States, [email protected] 1 - Partial Complementarity, Strategic Decision and Performance of Firms in Decentralized Supply Chains Xiang Fang, Case Western Reserve University, Cleveland, OH, 44106, United States, [email protected], Yunzeng Wang
Consider multiple manufacturers each producing a different product. The manufacturers rely on each other to fulfill a common demand stream from a type of customers who buy the multiple products as complementary sets. In addition, each manufacturer faces a separate customer type who buys only the manufacturer’s own product. All streams of demands are uncertain. We explore the equilibrium behavior of manufacturers in choosing their individual production capacities or quantities.

2 - Real-Time Optimization with ROMeo & IPOPT Brian Baumrucker, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected], Lorenz Biegler
The ROMeo (Rigorous On-line Modeling with Equation based Optimization) process modeling environment incorporates both reduced space SQP and IPOPT algorithms and is modified to deal with complementarity constraints. Numerical comparisons of the solvers will be presented and complementarity reformulations will be explored on a number of test problems.

3 - Circuit Tuning with IPOPT Andreas Waechter, Research Staff Member, IBM TJ Watson Research Center, 1101 Kitchawan Road, Yorktown Heights, NY, 10598, United States, [email protected]
During the design of digital circuits, such as microprocessors, the optimal transistor widths have to determined, in order to achieve fast performance, given power and area constraints. We will describe the implementation of the IBMinternal EinsTuner circuit tuning tool, which uses circuit simulations and the IPOPT nonlinear optimizer.

I SB03
Applications of Linear Programming
Sponsor: Optimization/ Linear Programming and Complementarity Sponsored Session
Chair: Javier Pena, Associate Professor of Operations Research, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected] 1 - A Robust CVaR Portfolio Allocation Model Using Option Prices Luis Zuluaga, Assistant Professor, University of New Brunswick, PO Box 4400, Fredericton, NB, E3B 5A3, Canada, [email protected], Juan Vera, Javier Pena
We present a novel linear programming portfolio allocation model in which worst-case Conditional Value-at-Risk (CVaR) is used as the risk measure. The worst-case CVaR is computed by considering underlying asset price probability distributions that replicate the observed prices of vanilla options.

2 - The Effect of Price Fluctuations on the Performance of Distribution Supply Chains Srinagesh Gavirneni, Assistant Professor of Operations Management, Johnson Graduate School of Management, Cornell University, 310 Sage Hall, Ithaca, NY, 14853, United States, [email protected], Shu (Vickii) Zhou
We consider a decentralized distribution supply chain with one supplier and two retailers. The supplier provides one product to the retailers, and each retailer faces iid end-customer demands. We want to identify the effect of price fluctuations and information sharing on the supply chain performance in terms of the total supply chain costs. We also want to determine the optimal magnitudes of the price fluctuations and schedule the price fluctuations across retailers to maximize the benefit.

3 - Monotone Properties of Wholesale Price Contracts Vera Tilson, Asssistant Professor, University of Rochester, Simon School of Business, Rochester, NY, United States, [email protected]
We focus on monotonicity properties of wholesale price — performing qualitative sensitivity analysis using order-theoretical framework. We establish sufficient conditions that guarantee monotone changes in contract parameters with changes in exogenous parameters. Many specific cases are encompassed by the assumptions including randomness in both yield and demand.

2 - Static Arbitrage Bounds for Basket Options via Linear Programming Juan Vera, Postdoctoral Fellow, Georgia State University, United States, [email protected], Luis Zuluaga, Javier Pena
We show that the problem of finding the best possible bounds on a basket option given the prices of other basket options can be cast as a linear program. We subsequently show that for special cases, e.g., when only calls on individual assets are given, the linear programming formulation can be simplified substantially and has a closed-form solution.

I SB05
Optimization Approaches in Financial Engineering I
Sponsor: Optimization/ Stochastic Programming Sponsored Session
Chair: Stan Uryasev, Professor, University of Florida, Department of Industrial and Systems Engineering, Gainesville, FL, 32601, United States, [email protected] 1 - How Much Is a Model Upgrade Worth? Sven Sandow, Director, Quantitative Analytics, Standard & Poor’s, 55 Water Street, 46th Floor, New York, NY, 10041, United States, [email protected], James Huang, Craig Friedman
We compute, from the point of view of an expected-utility-maximizing investor, the monetary values of probabilistic models that are used for financial decisionmaking. In the case of nearly homogeneous returns, the value of a model upgrade can be approximated by a function of the likelihood ratio.

3 - A Gradient-Based Method for Computing Nash Equilibria of Two-Person Zero-Sum Sequential Games Samid Hoda, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected], Javier Pena, Andrew Gilpin
We discuss an adaptation of Nesterov’s smoothing technique to compute approximate equilibrium points of two-person, zero-sum sequential games. The algorithm is gradient-based with a rate of convergence O(1/k) where k is the iteration count, but with fairly low computational cost per iteration. The latter makes it particularly suitable to find approximate Nash equilibria of large games.

4 - Alternative Transportation Fuels: Modeling National-Scale Ethanol Production and Distribution Scott Matthews, Associate Professor, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, 15213, United States, [email protected], Michael Griffin, William R. Morrow
Crude oil refining and gasoline distribution have been optimized over time. Ethanol is currently a niche market. Our LP model estimates the logistics costs associated with national ethanol production and distribution from biomass. We estimate higher downstream transportation costs than gasoline (15-48 cents compared to 3 cents per gallon) but remain a fraction of total cost at blends up to 16% ethanol. Total infrastructure constraints could make large-scale ethanol distribution infeasible.

2 - Methods of Reducing Optimization of Omega Function to Linear Programming Sergey Sarykalin, PhD Student, University of Florida, 303 Weil Hall, PO Box 116595, Gainesville, FL, 32611, United States, [email protected], Stan Uryasev
We consider the problem of maximizing the Omega Function of a portfolio with a fixed benchmark. We propose two theorems reducing the original problem to the two relaxed problems without a restriction that the decision variables should sum up to 1. When the Omega function at optimality is greater than 1, both relaxed problem can be reduced to linear programming. We illustrate the proposed methodology with the case study on allocating recourses among managers in a hedge fund.

69

SB06

INFORMS PITTSBURGH — 2006 I SB07
New Books in Optimization
Cluster: New Books Invited Session
Chair: Stephen Wright, Professor, University of Wisconsin-Madison, 1210 West Dayton Street, Madison, WI, 53706, United States, [email protected] 1 - Interior Point Approach to Linear Optimization: Theory and Algorithms Tamás Terlaky, Professor, McMaster University, 1280 Main Street West, Hamilton, ON, L8S4K1, Canada, [email protected], Kees Roos, Jean-Philippe Vial
An interior point approach to both theoretical and algorithmic aspects of interior point methods for linear optimization. The first part presents duality theory and a simple polynomial IPM with full complexity and roundig analysis for solving linear optimization problems. Remainder discusses logarithmic barrier methods, target-following approaches, predictor-corrector approaches, sensitity analysis and implementation issues.

3 - Default Hazard Rate Modeling with Evolving and Non-Evolving Explanatory Variables Craig Friedman, Standard & Poor’s, 55 Water Street, New York, NY, 10041, United States, [email protected]
We discuss a statistical learning approach to default intensity (hazard rate) modeling. This approach allows us to incorporate explanatory variables whose evolution we want to model (such as common macroeconomic default drivers), as well as informative variables that we may not want to model (such as firmspecific financial ratios)

I SB06
Novel Randomization Approaches in Search
Sponsor: INFORMS Computing Society/Constraint Programming Sponsored Session
Chair: Carla Gomes, Associate Professor, Cornell University, 4130 Upson Hall, Ithaca, NY, United States, [email protected] 1 - Disco-Novo-GoGo: Integrating Local Search and Complete Search with Restarts Meinolf Sellmann, Assistant Professor, Brown University, 115 Waterman Street, PO Box 1910, Providence, RI, 02912, United States, [email protected], Carlos Ansotegui, Warren Schudy
A hybrid algorithm is devised to boost the performance of complete search on under-constrained problems. We suggest to use random variable selection in combination with restarts, augmented by a coarse-grained local search algorithm that learns favorable value heuristics over the course of several restarts. Numerical results show that this method can speed-up complete search by orders of magnitude.

2 - Applied Optimization: Formulation and Algorithms for Engineering Systems Ross Baldick, Professor, The University of Texas, 1 University Station, Austin, TX, 78712, United States, [email protected]
The motivation for this book stems from my observations of the orientation of typical optimization texts used in optimization courses compared to the needs of students in our engineering program. Many optimization books and courses concentrate on the design of algorithms. In contrast, this book focuses on how to formulate simultaneous equations and optimization problems so that they can be solved by existing software.

2 - The Max K-Armed Bandit Problem: Mediating the Use of Multiple Search Heuristics Stephen Smith, Research Professor, Carnegie Mellon University, 5000 Forbes Avenue, The Robotics Institute, Pittsburgh, PA, 15213, United States, [email protected]
Randomization has been shown to be an effective means of boosting the performance of search heuristics in combinatorial domains, yet it is often the case that the utility of a given heuristic varies across problem instances. In this talk we consider the problem of allocating trials among a set of heuristics to best exploit their differential power, which we formalize as the Max k-armed bandit problem. We summarize recent theoretical results and show their application to scheduling problems.

3 - Numerical Optimization, Second Edition, J. Nocedal and S. Wright Stephen Wright, Professor, University of Wisconsin-Madison, 1210 West Dayton Street, Madison, WI, 53706, United States, [email protected], Jorge Nocedal
Additions and improvements in the second edition of the book “Numerical Optimization” by J. Nocedal and S. Wright will be discussed. An outline of the contents and a sample of the approach used in the book will be presented.

3 - Supporting Parallel and Distributed Constraint-Based Local Search Algorithms. Laurent Michel, Assistant Professor, Computer Science & Engineering Department, University of Connecticut, Storrs, CT, 06269, United States, [email protected], Pascal Van Hentenryck
Optimization applications are challenging because of size, complexity or online requirements. At the same time, commodity multiprocessors offer significant opportunities but remain extremely challenging to program and use. It is therefore capital to lower the barrier to their adoption in the optimization domain. This talk offers an answer: programming language abstractions to minimize the distance between sequential and parallel and distributed algorithms.

4 - Nonlinear Optimization Andrzej Ruszczynski, Professor, Department of Management Science and Information Systems, Rutgers University, 94 Rockefeller Road, Piscataway, NJ, 00854, United States, [email protected]
The book covers convex analysis, optimality conditions, duality, and methods for solving unconstrained and constrained problems. It addresses modern topics such as optimality conditions and numerical methods for problems involving nondifferentiable functions, semidefinite programming, metric regularity and stability theory of set-constrained systems, and sensitivity analysis of optimization problems. All results are proved and illustrated by numerous examples and figures.

I SB08
Behavioral Operations
Sponsor: Technology Management Sponsored Session
Chair: Gary Pisano, Professor of Business Administration, Harvard Business School, Morgan Hall 417, Boston, MA, 02163, United States, [email protected] Co-Chair: Francesca Gino, Lecturer and Post-Doctoral Fellow, Harvard Business School, Morgan Hall 412, Boston, MA, 02163, United States, [email protected] 1 - Cross-Functional Coordination in Supply Chain (SC) Forecasting and Planning Noel Watson, Assistant Professor, Harvard Business School, Soldiers Field Road, Medford, MA, 02155, United States, [email protected]
Quant. models of cross-functional coordination primarily exploit incentives which require complete system-wide knowledge by a planner. In practice, some efforts have instead focused on specific process mechanisms such as improving directly the collective information processing abilities of participating functions. We re-examine then coordination to see what is achievable without primarily changing incentives. Behavioral dynamic constructs such as perceptions are valuable in our models.

4 - Streamlining Reasoning for Solution Finding and Counting Carla Gomes, Associate Professor, Cornell University, 4130 Upson Hall, Ithaca, NY, United States, [email protected], Ashish Sabharwal, Bart Selman, Meinolf Sellmann
We will introduce streamlining reasoning, a technique for effective solution finding and solution counting for hard combinatorial problems.

70

INFORMS PITTSBURGH— 2006
2 - A Behavioral Investigation of Service-Based Supplier Competitions Karen Donohue, Associate Professor, Carlson School, University of Minnesota, 321 19th Avenue S., Minneapolis, MN, 55455, United States, [email protected], Saif Benjaafar, Elena Katok
We compare two types of supplier competitions that are proven, in theory, to be equally effective ways for a buyer to elicit high service quality from his supply base. We study the two mechanisms in the laboratory to determine whether these similarities hold true with human decision-makers. We also test the effect of the number of competitors and different service cost structures on service performance.

SB11

3 - An Empirical Analysis of R&D Portfolios in the Pharmaceutical Industry Christian Terwiesch, Associate Professor, University of Pennsylvania, 548 JMHH, Philadelphia, PA, 19104, United States, [email protected], Karl Ulrich, Karan Girotra
We present an empirical analysis of R&D performance in the pharmaceutical industry. Our focus is on the link between product portfolio and financial performance. We also explore various dimensions of R&D performance, including the R&D yield and the R&D efficiency.

3 - Bridging Laboratory and Field Behavior: Bargaining and Auctions in Controlled Experiments on EBay Gary Bolton, Smeal College of Business, Penn State University, 334 Business Building, University Park, PA, United States, [email protected], Axel Ockenfels
We conducted a controlled field experiment on eBay to investigate first, whether basic laboratory results on equitable bargaining and competitive bidding surface in a naturally occurring market environment among experienced traders, and second, whether trading strategies observed in the experiment can be linked to trading patterns observed outside the experiment.

I SB10
Software Demonstration
Cluster: Software Demonstration Invited Session
1 - Dash Optimization - Building Optimization Applications Using Mosel and XAD Nitin Verma, Dash Optimization, 560 Sylvan Avenue, Englewood Cliffs, NJ, United States, [email protected]
This tutorial focuses on building optimization models using simple and advanced features available in Mosel, our modeling and programming language, and integrating them seamlessly with Xpress-Application Developer for visualization, analysis, and deployment purposes. For updated information about Dash Optimization’s activities meeting, please visit: www.dashoptimization.com/home/services/news/informs_nov2006.html.

4 - Research Proposal: The Cognitive Aspects of Scheduling Yishai Boasson, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA, 02139, United States, [email protected]
We wish to understand how expert schedulers modify scheduling heuristics and choose between them. Field research methods will be used to collect data about expert scheduler behavior. Data will be compared and contrasted with lab data. Insights gained can be used in models for system design. This is different from previous works, as it is concerned with how schedulers make decisions, rather than with what those decisions are, or with benchmarking scheduler performance against normative theory.

2- Rapid Application Prototyping with GAMS Steven Dirkse,GAMS Development, 1217 Potomac Street NW, Washington DC 20007, United States, [email protected]
GAMS Development will demonstrate how an application can be built using GAMS. We’ll use both fundamental modeling practices, our state of the art solvers, and the latest in data access and application integration tools to quickly produce a working application: a Sudoku calculator sure to impress friends and family!

5 - Experimental Operations Research J. Neil Bearden, Assistant Professor, University of Arizona, Eller College of Management, Mcclelland Hall 417, Tucson, AZ, 85720, United States, [email protected]
I survey laboratory studies of decision behavior in problems drawn from theoretical OR literature, focusing on job sequencing and revenue management. I argue that we should not be concerned with the “irrationality” of human decision making, but use OR models as standards to evaluate where decision making can be improved.

I SB11
OR in SpORts I
Sponsor: OR in Sports Sponsored Session
Chair: Eric Huggins, Assistant Professor of Management, Fort Lewis College, 1000 Rim Drive, Durango, CO, 81301, United States, [email protected] 1 - Does the Pitcher or the Batter Control Home Plate in a Major League Baseball Game? James Cochran, Louisiana Tech University, College of Administration & Business, PO Box 10318, Ruston, LA, 71055, United States, [email protected]
In baseball the question of who has more influence over the outcome of a plate appearance is critical - the answer may help teams assess value of players and decide at which positions to invest. We use regression analysis and data from the 04 MLB season - frequency of particular events (hits, homeruns, etc.) for individual pitcher/batter matchups and the frequency of the event for individual pitcher and hitters over the entire season - to assess relative influence of the pitcher and the batter.

I SB09
Empirical and Experimental Research in New Product Development
Cluster: New Product Development Invited Session
Chair: Kamalini Ramdas, Associate Professor, Darden School, 189 FOB, 100 Darden Boulevard, Charlottesville, VA, 22902, United States, [email protected] 1 - The Lone Inventor: Kook or Hero? Lee Fleming, Professor, Harvard Business School, 485 Morgan Hall, Boston, MA, 02163, United States, [email protected]
Why are lone inventors thought to be the sources of technological breakthroughs? The myth persists despite a variety of arguments that collaboration increases creativity. I argue and demonstrate that while the average rate and success of creative effort is lower for lone inventors, the successes are much more variable. If assessments of success depend on the maximum of a distribution, then a higher variance distribution could be a plausible basis for the “myth.”

2 - A Player Selection Heuristic for a Sports League Draft Michael Fry, Assistant Professor, University of Cincinnati, 532 Lindner Hall, QAOM Department, Cincinnati, OH, 45221-0130, United States, [email protected], Jeffrey Ohlmann, Andy Lundberg
We model the decision-making process of a sports franchise during a player selection draft. Our dynamic programming-based model considers both the needs of a particular team as well as the overall abilities of the players available to be drafted. Our model can be solved in real time during a draft using a spreadsheetbased decision support system. We provide an example of our strategy applied to a fantasy football draft and compare our proposed solution strategy to several competing heuristics.

2 - Social Preferences and Performance in Buyer-Supplier Relationships Christoph Loch, Professor of Technology Management, INSEAD, Boulevard de Constance, Fontainebleau, France, [email protected], Yaozhong Wu
Economically rational buyer-supplier partners need incentive contracts in order to efficiently coordinate. However, social preferences, or concerns for status, fairness and relationships, can allow parties to collaborate without formal contracts. We show in a model that social preferences can lead to successful coordination, and we show experimental evidence with human subjects that manipulating social concerns in subjects indeed reliably influences their behavior in a buyer-supplier setting.

3 - Optimal NFL Survivor Eric Huggins, Assistant Professor of Management, Fort Lewis College, 1000 Rim Drive, Durango, CO, 81301, United States, [email protected]
NFL Survivor is a game where players try to choose one winning team each week for the 17 week NFL season, without choosing the same team twice. Players that pick a losing team are eliminated until one winner (survivor) remains. We show that a greedy algorithm, picking the best team each week, is near (but not necessarily) optimal and discuss heuristic strategies. We also discuss results for a generalized version of the game.

71

SB12 I SB12

INFORMS PITTSBURGH — 2006
2 - Inspection and Quality Improvement in Manufacturing Systems Jeroen Vits, Research Group Operations Management, Katholieke Universiteit Leuven, Celestijnenlaan 300A, Heverlee, 3001, Belgium, [email protected], Philippe Chevalier
We characterize optimal inspection and quality improvement policies for single stage systems and compute inspection and quality improvement thresholds in the presence of imperfect inspection, forgetting and natural learning. For multi-stage systems, we present and evaluate performance of heuristic un-interrupted inspection and quality improvement policies and give thresholds for inspection and quality improvement.

2006 Dantzig Dissertation Award Finalists
Cluster: Dantzig Dissertation Prize Invited Session
Chair: Robert Shumsky, Associate Professor, Tuck School, Dartmouth College, 100 Tuck Hall, Hanover, NH, 03755, United States, [email protected] 1 - Heterogeneous Congestion Control Protocols A. Kevin Tang, Caltech, MC 256-80, Pasadena, CA, 91125, United States, [email protected]
Homogeneity of price is an implicit yet fundamental assumption underlying price based resource allocation theory. In this thesis, we study effects of relaxing this assumption by examining networks with heterogeneous congestion control. A systematic theory is developed that includes all major properties of equilibrium such as existence, uniqueness, optimality and stability. Furthermore, a practical distributed algorithm is proposed to steer a network to the unique optimal operating point. Various experiments are provided to illustrate the theory and verify its predictions.

3 - Closed-Loop System Identification for Small Samples with Constraints Arda Vanli, PhD Student, The Pennsylvania State University, 350 Leonhard Building, University Park, PA, 16802, United States, [email protected], Enrique Del Castillo
The traditional approaches to the closed-loop system identification problem require using sufficiently large data sets and general enough model forms while at the same time applying some form of a dither signal to the process. We propose an identification procedure to improve model parameter estimates by incorporating prior knowledge about the process in the form of constraints without the use of a dither signal. A Monte Carlo simulation study is presented to illustrate the method.

2 - R&D Investment Dynamics under Uncertainty: Theory and Application Tao Yao, Assistant Professor, The Pennsylvania State University, 349 Leonhard Building, University Park, PA, 16802, United States, [email protected]
Strategic R&D Investment under Uncertainty: Collusion and Information Time Lag. Introduce collusion equilibrium under the repeated strategic interactions. Explain several otherwise puzzling innovation market phenomena. Analyze the likelihood of tacit collusion in R&D and note implications for antitrust and merger control. Dynamic R&D Project Selection. Project selection decisions under uncertainty are often made sequentially. Use repeated options to derive optimal rules. Illustrate with a Hybrid Electric Cars vs. Hydrogen Fuel Cell Vehicles example as pursued by GM.

4 - Bayesian Modeling of Call Center Arrivals Refik Soyer, The George Washington University, Funger Hall 415, Decision Sciences, 2201 G Street, NW, Washington, DC, 20052, United States, [email protected], Murat Tarimcilar
A modulated Poisson process model is presented to describe arrivals to a call center. The model takes into account both covariate and time effects on call volume intensity and enables us to assess effectiveness of different advertising strategies along with predicting the arrival patterns. A Bayesian analysis of the model is developed and an extension is presented to consider heterogeneity in arrivals. The model and the methodology are implemented using real call center arrival data.

I SB13
Teaching Case Competition II
Sponsor: Education (INFORM-ED) Sponsored Session
Chair: Tasha R. Inniss, Assistant Professor of Mathematics, Spelman College, 350 Spelman Lane, SW, Box 320, Atlanta, GA, 30126, United States, [email protected] 1 - Presentations of Finalists #3 and #4 for the Seventh Annual INFORMS Case Competition Tasha R. Inniss, Assistant Professor of Mathematics, Spelman College, 350 Spelman Lane, SW, Box 320, Atlanta, GA, 30126, United States, [email protected]
The FOUR finalists for the 2006 INFORMS Case Competition will deliver final presentations of their teaching cases to a panel of pre-selected judges and the audience. All are welcome to attend and observe these presentations as well as ask questions of the finalists. At the end of the four presentations, the judges will select the winner of the Case Competition. The winners and runners-up will be announced and presented with plaques as well as cash awards at the INFORMED Annual Business Meeting.

5 - Sigma Six as a Quality Control Method Syed Shahabuddin, Professor, Central Michigan University, 1306 E. Preston, Mount Pleasant, MI, 48859, United States, [email protected]
Sigma Six has been touted as a method to use for quality control. Many researchers and practitioners believe that Sigma Six should be a tool of quality control. The search of the literature shows companies claiming that the use of Sigma Six has saved money and has improved customers’ satisfaction. The question many should be asking, however: Is this practical and feasible? My paper will discuss the concept of Sigma Six and the feasibility of its application in real business conditions.

I SB15
Procurement and Combinatorial Auctions I
Cluster: Auctions and e-Commerce Invited Session
Chair: Aleksandar Pekec, Associate Professor, Duke University, Fuqua School of Business, 1 Towerview Drive, Durham, NC, 27708, United States, [email protected] 1 - Optimal Procurement Auctions of Divisible Goods with Capacitated Suppliers Anuj Kumar, Industrial Engineering & Operations Research Department, Columbia University, 500 West 120th Street, New York, NY, 10027, United States, [email protected], Garud Iyengar
Most of the literature, in procurement auctions (reverse auctions) assumes uncapacitated suppliers. Consequently, in a constant marginal production cost model, these auction mechanisms award the contract to a single supplier. We consider a model with capacitated suppliers where, in addition to the privately known marginal costs, the production capacities are also private information. We construct the optimal direct mechanism that maximizes the retailers expected profits in this model. The solution is closed form under {\it regularity}. The results in this paper are applicable to a number of principle-agent mechanism design problems (for example monopoly pricing with adverse selection, forward auctions) where the agents have privately known upper bound on the allocations. The sealed low bid implementation of the optimal direct mechanism is also presented in symmetric environment. The implementation requires that each supplier submit a bid consisting of the marginal payment sought and total capacity available. These bids serve as the input to a simple optimization problem that computes the quantity allocation for each firm. We extend the model to multi-product procurement with complementarities.

I SB14
Quality Control
Contributed Session
Chair: Syed Shahabuddin, Professor, Central Michigan University, 1306 E. Preston, Mount Pleasant, MI, 48859, United States, [email protected] 1 - Design Considerations and Analysis Of Multistage Split-Plot Nanomanufacturing Experiments Navin Acharya, PhD Candidate, Pennsylvania State University, 445 Waupelani Drive, Apt D-16, State College, PA, 16801, United States, [email protected], Harriett Nembhard
Many processes in nanomanufacturing are conducted over a series of stages. Specifically, a polymerization process for the fabrication of nano-films is studied. To execute efficient experimentation and fully understand the intricacies at the nano scale, split-plot designs that can be applied effectively over multiple stages are proposed and their characteristics examined. General expressions for some of the properties of these designs and analysis are developed.

72

INFORMS PITTSBURGH— 2006
2 - Automated Design of Multistage Mechanisms Tuomas Sandholm, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, United States, [email protected], Vincent Conitzer, Craig Boutilier
We extend automated mechanism design to the design of multistage mechanisms. Our approach is based on algorithms for converting single-stage dominant-strategy mechanisms into multistage mechanism. For Bayes-Nash mechanisms, we prove that the preference elicitor may be better off randomizing over queries and hiding elicited results from the agents.

SB17

I SB17
Revenue Management Models
Contributed Session
Chair: Craig Sorochuk, University of Western Ontario, Richard Ivey School of Business, 1151 Richmond Street North, London, ON, N6A 3K7, Canada, [email protected] 1 - Optimal Pricing of Finite Capacity Amit Eynan, Professor, University of Richmond, School of Business, Richmond, VA, 23173, United States, [email protected], Chakravarthi Narasimhan
A firm owns finite capacity for the delivery of services or production of customized products. When a potential buyer indicates the amount of capacity he would like to employ the firm suggests a price while being challenged by lack of exact information regarding buyer’s reservation price (only its distribution is known), and future requests from additional potential buyers. Managerial insight regarding the optimal price and potential profit increase are provided.

3 - An Evaluation of Competitive Bids in Drywall Construction Trades Gerald H. Williams, R. Brown Consulting Group, Construction & Engineering Management Research, Inc., 2300 SW First Avenue, Suite 102, Portland, OR, 97212-5047
This paper presents the final results of a study comparing the estimated and actual labor productivity for drywall trades on 400 construction projects in the western US. We document a wide variance in estimated productivities in these trades and attempt to account for variance in actual labor productivities.

4 - Auction Theory in OR/MS Settings Aleksandar Pekec, Associate Professor, Duke University, Fuqua School of Business, 1 Towerview Drive, Durham, NC, 27708, United States, [email protected]
Standard auction theory results are sensitive to assumptions important in MS and OR applications. Uncertainty in supply/demand has implications on equilibrium existence and revenue rankings of uniform and discriminatory multi-item auctions. Furthermore, choice of an auction information policy and supply quantities are important parameters for revenue optimization.

2 - Revenue Management with Multiple Capacity Dimensions Wei Yang, Assistant Professor, Long Island University at C.W. Post, 720 Northern Boulevard, Brookville, NY, 11548, United States, [email protected], Baichun Xiao
In container shipping, air cargo management, trucking and health care industry products or services possess multiple capacity attributes. We formulate the revenue management problem with multiple capacity features into a continuoustime stochastic control model. We derive the optimal solution in closed form and show that considering both price and demand intensity the optimal policy is of threshold and displays a significant difference when the remaining capacity-mix varies.

I SB16
Topics in Revenue Management and Pricing
Cluster: Dynamic Pricing and Forecasting Invited Session
Chair: Andrew Boyd, Chief Scientist and Senior Vice President Science and Research PROS, 3100 Main Street, Suite 900, Houston, TX, 77002, United States, [email protected] 1 - Management Issues in Pricing and Revenue Optimization Brenda Barnes, President, Barnes Consulting Company, 140 E. 28th Street, Suite 10H, New York, NY, 10016, United States, [email protected]
Pricing and revenue optimization systems can increase a firm’s profitability by 8 to 10 percent when properly implemented and managed. However, a company’s results will vary due to environment-specific factors impacting system performance, such as data integrity, analyst intervention, lack of coordination, and operational disruption. We draw on industry experience to identify factors impacting system effectiveness and actions management should take to achieve maximum financial benefits.

3 - Revisiting Robustness of Capacity Rationing Models Mehmet Barut, Wichita State University, 1845 Fairmount, 324 Clinton Hall, Wichita, KS, 67260-0077, United States, [email protected], Mehmet Bayram Yildirim, V Sridharan
The goal of this study is to investigate the power and generalizeability of the existing stochastic models, developed for capacitated order management in MTO environment; and to see if these models can be used as a tool for strategic exit. The study uses a locked two-step process: a simulation model and a mathematical model; relaxes the assumption of decreasing demand over time; and provides statistical evidence for the performance of current models under variety of demand patterns.

4 - Fencing in the Context of Revenue Management Michael Zhang, Saint Mary’s University, 903 Robie Street, Halifax, NS, B3H1Y5, Canada, [email protected]
Market segmentation is a key strategic element in the practice of revenue management. After being identified, market segments should be kept separate to prevent demand spillover from high priced segments to low priced segments. Tools to restrict customer migration across segments are referred as fences. The purpose of this study is to address managerial implication for RM from a fencing perspective. The pertinent issues surround fences are explored.

2 - A Monopoly’s Location and Spatial Pricing Strategies Binbin Liu, PhD Candidate, Rotman School of Management, University of Toronto, 105 St. George Street, Toronto, ON, M5S 3E6, Canada, [email protected], Dmitry Krass, Oded Berman
A multi-store monopoly’s simultaneous pricing and location decisions on networks are studied. Three pricing policies, discriminatory, uniform and mill pricing are considered. We establish nodal optimality of the location decision and find optimal locations and prices under mild conditions. We provide managerial insights regarding when a firm should use discriminatory pricing versus mill or uniform pricing.

5 - Revenue Management in Sports Craig Sorochuk, University of Western Ontario, Richard Ivey School of Business, 1151 Richmond Street North, London, ON, N6A 3K7, Canada, [email protected], John Wilson
While revenue management has been well-studied and applied to the airline and hotel industries for years, its applications in sports industries are relatively sparse. Given that the four major sports leagues in North America (MLB, NBA, NFL and NHL) comprise a $10 billion+ industry, there are certainly opportunities for implementation of revenue management techniques. This presentation includes a literature survey and outlines potential avenues of research of revenue management in sports.

3 - The Value of Attaching the Correct Revenue to an Airline Booking Igor Cakulev, Senior Development Scientist, PROS, 3100 Main Street, Suite 900, Houston, TX, 77002, United States, [email protected], Andrew Boyd
As a result of the way airline tickets are sold, carriers frequently use an approximation of the revenue they will pocket when they sell a ticket. This raises the question: what is the value of using the exact revenue? The business problem is described and ways to overcome it are discussed, and results are presented showing the achievable revenue improvements.

73

SB18 I SB18
Models of Energy Markets

INFORMS PITTSBURGH — 2006
2 - Multi-Objective Optimization as an Aid to Sell Non-Timber Forest Benefits in New Zealand Sandor Toth, Assistant Professor, Oklahoma State University, Stillwater, OK, 74078, United States, [email protected]
Many argue that developing a capability to value and market the environmental benefits of forests could alleviate some of the problems associated with forest degradation and land-use conversions. Multi-objective harvest scheduling models can help landowners sell non-timber forest benefits by quantifying the tradeoffs and opportunity costs of producing these services. We will demonstrate how biodiversity and carbon credits could be marketed using this approach in New Zealand.

Sponsor: Energy, Natural Resources & The Environment Sponsored Session
Chair: David Fuller, Professor, Department of Management Sciences, University of Waterloo, 200 University Avenue W., Waterloo, ON, N2L 3G1, Canada, [email protected] 1 - Finding Supply Function Equilibrium in Electricity Markets using an AC Approach Miguel Anjos, Assistant Professor, University of Waterloo, 200 University Avenue West, Waterloo, ON, N2L 3G1, Canada, [email protected], Anthony Vannelli, Guillermo Bautista
Using an AC transmission network, oligopolistic competition in power markets is formulated as a Nonlinear Programming (NLP) problem, and characterized by a multi-leader single-follower game. The transmission network is modelled with a detailed nonlinear system. For the game setting, a supply function equilibrium approach is used.

I SB20
Portfolio Decision Analysis
Sponsor: Decision Analysis Society Sponsored Session
Chair: Jeffrey Keisler, Assistant Professor, University of Massachusetts Boston, MSIS Department, 100 Morrissey Boulevard, Boston, MA, 02125, United States, [email protected] 1 - Participatory Priority-Setting Processes with Robust Portfolio Modeling Ahti Salo, Professor, Helsinki University of Technology, Systems Analysis Laboratory, PO Box 1100, Otakaari 1 M, Espoo, 02015, Finland, [email protected]
The collaborative development of research priorities typically involves high uncertainties, multiple objectives and a large number of participants. We report case studies where internet-based decision aiding tools, Robust Portfolio Modelling and participatory workshops have served to inform the development of national priorities. We also discuss network-oriented portfolio decision analyses, with examples from an international consultation process involving over 400 participants from 8 countries.

2 - Models of Efficient Pricing Schemes in Electricity Markets Emre Celebi, Department of Management Sciences, University of Waterloo, 200 University Avenue West, Waterloo, ON, N2L3G1, Canada, [email protected], David Fuller
This presentation will examine the mixed complementarity programming models of equilibrium in competitive electricity markets that can bridge the speed of response gap between suppliers and consumers, yet adhere to the principle of marginal cost pricing. A computable equilibrium model is developed to estimate the time-of-use (TOU) prices that can be used in retail electricity markets. Also strategic interactions among suppliers are examined within this model (e.g., Cournot, Stackelberg).

3 - Variational Inequality Formulations of Energy Market Models: A Decomposition Approach David Fuller, Professor, Department of Management Sciences, University of Waterloo, 200 University Avenue W., Waterloo, ON, N2L 3G1, Canada, [email protected], William Chung
We outline a new decomposition method for variational inequalities that represent a generalized Nash equilibrium, i.e., with constraints that link the Nash players. The subproblems are ordinary optimization models, which, for many energy market models, are the starting point of the modelling process; this observation simplifies the process of defining the whole VI model. We illustrate with a Nash-Cournot model of an electricity market.

2 - Constructing Organisational Preferences: Within-Criterion Weighting in Multi-Criteria Portfolio Alec Morton, [email protected], Lawrence Phillips
Decision Analysis often helps people discuss what precisely constitutes value for their organisation. Decision Conferencing in particular uses multiple workshops with multiple areas of the organization to develop a value model. Locally assessed valuations may require “within-criterion weights” to arrive at a global value scale. We shall interpret these weights, and explore their ability to reduce error, and the appropriateness of different ways of structuring an intervention which uses them.

4 - A Day Ahead Electricity Market with Reserve Shortage Pricing Jichen Zhang, Department of Management Sciences, University of Waterloo, 200 University Avenue West, Waterloo, ON, N2L3G1, Canada, [email protected], Samir Elhedhli, David Fuller
We present a large-scale multi-period stochastic linear programming model for scheduling the electric energy in a day-ahead electricity market. The model considers various scenarios and integrates the idea of reserve shortage pricing based on reserve shortage level in real time. Numerical results are discussed.

3 - Analyzing a Portfolio Decision Process Jeffrey Keisler, Assistant Professor, University of Massachusetts Boston, MSIS Department, 100 Morrissey Boulevard, Boston, MA, 02125, United States, [email protected], Jeffrey Stonebraker
We obtained summary data from prior decision analyses of the entire new product pipeline at a major pharmaceutical company. These data include success probabilities for different phases of development, and estimates of cost and market value, coupled with information about project maturity, therapeutic area, etc. We analyze this data to characterize the portfolio and identify where high quality estimates are critical. This yields recommendations for decision process improvement.

I SB19
Forestry Applications II
Sponsor: Energy, Natural Resources & The Environment Sponsored Session
Chair: Robert Haight, Research Forester, USDA Forest Service, North Central Research Station, 1992 Folwell Avenue, St. Paul, MN, 55108, United States, [email protected] 1 - Using Markowitz Portfolio Optimization to Guide Reforestation Under Multiple Climate Change Scenario Kevin Crowe, Assistant Professor, Faculty of Forestry, Lakehead University, 955 Oliver Road, Thunder Bay, ON, P7B 5, Canada, [email protected], William Parker
A major difficulty of planned adaptation to climate change arises from the uncertainty of estimated changes in climate parameters. The general problem addressed by this study is that of designing a decision support system to select an optimal set of seed sources to be used in regenerating forests in an environment of multiple, equally plausible future climates.

I SB21
Ideas Relevant to Multiattribute Decision Analysis Practice
Sponsor: Decision Analysis Society Sponsored Session
Chair: Ralph Keeney, Research Professor, Fuqua School of Business, Duke University, Box 90120, Durham, NC, 27708-0120, United States, [email protected] 1 - The Role of Multi-Criteria Decision Analysis in Helping the Committee on Radioactive Waste Lawrence Phillips, Professor, London School of Economics & Political Science, Department of Operational Research, Houghton Street, London, NW3 1AH, United Kingdom, [email protected]
This paper reports on the largest public consultation exercise ever undertaken in the United Kingdom, with emphasis on the role of multi-criteria decision analysis (MCDA) in helping the Committee on Radioactive Waste Management (CoRWM) to recommend how the country’s accumulating high-level and medium-level radioactive waste should be managed.

74

INFORMS PITTSBURGH— 2006
2 - Bidirectional Utility Diagrams Ali Abbas, Assistant Professor, College of Engineering, University of Illinois-Urbana Champaign, 104 S. Mathews Avenue, Urbana, IL, 61822, United States, [email protected]
We present a graphical representation of multiattribute utility functions that captures the asymmetric nature of utility dependence. This representation facilitates the elicitation of multiattribute utility functions and enables their decomposition into lower order terms that are easier to elicit by knowledge of utility independence relations among the attributes.

SB25

I SB24
Cultural Pragmatics in Professions
Sponsor: Organization Science Sponsored Session
Chair: Klaus Weber, Assistant Professor of Management and Organizations, Kellogg School of Management, Northwestern University, 2001 Sheridan Rd, Evanston, IL, 60208-2001, United States, [email protected] Co-Chair: Amit Nigam, Post Doctoral Researcher, Rotman School of Management, University of Toronto, 105 St. George Street, Toronto, ON, M5S 3E6, Canada, [email protected]
The papers in this session examine the cultural pragmatics, the skillful use of cultural elements for social outcomes, in the context of professional and occupational groups.

3 - Specifying Objectives for Decisions is not Easy Ralph Keeney, Research Professor, Fuqua School of Business, Duke University, Box 90120, Durham, NC, 27708-0120, United States, [email protected], Kurt Carlson, Samuel Bond
Our recent experiments indicate that many individuals cannot list an appropriate set of objectives for their own important decisions. Individuals identify only about half of their objectives. Furthermore, using their own priorities, they omit objectives as important as those they identify. Suggestions for improving the specification of objectives are given.

I SB22
Mission and Means Framework - Applications
Sponsor: Military Applications Sponsored Session
Chair: William Yeakel, President, ORSA Corporation, 1003 Old Philadelphia Road, Suite 103, Aberdeen, MD, 21001, United States, [email protected] 1 - Using the Missions and Means Framework to Determine Value Added of Systems of Systems Britt Bray, Senior Military Analyst, DRC, 106 S. 5th Street, Leavenworth, KS, 66048, United States, [email protected], Paul Tanenbaum
The MMF enables front end analysis of missions to determine required capabilities and measure effectiveness. The MMF helps analysts generate the most important questions to be answered. It provides a logical framework to store and understand the logical relationships between the data collected and document and understand the relationships between Systems of Systems (SOS) and the processes that employ them. This paper will describe how the MMF can be applied to determine value added of SOS.

1 - Redefining Professional Control: Clinical Guidelines, Quality Measures and the Scope for Professional Judgment Amit Nigam, Post Doctoral Researcher, Rotman School of Management, University of Toronto, 105 St. George Street, Toronto, ON, M5S 3E6, Canada, [email protected]
This paper looks at the changing nature of professional controls in a period of institutional change by focusing on changes in the medical profession in the shift to managed care in the United States. It shows that the medical profession shifts from defining quality based on the training and skills of physicians, towards defining it based on clinical guidelines and quality measures. It discusses the implications of each approach to quality for professional judgment and discretion.

2- Designing a Frame: Rhetorical Strategies of Architects for Client Engagement Candace Jones, Carroll School of Management, Boston College, Fulton Hall 510, 140 Commonwealth Avenue, Chestnut Hill, MA, 02467, United States, [email protected], Reut Livne-Tarandach
We examine how architectural firms framed their qualifications when competing for projects. Our exploratory analysis of firms’ rhetorics identified a limited number of key words, when framing their qualification for clients. Our exploratory analyses revealed that architectural firms used a severely limited number of key words to signal multiple competencies, design among them, and to resolve competing demands such as creativity, functionality and cost.

2 - Missions and Means Framework (MMF): Live-Fire Test and Evaluation Strategies Martha K. Nelson, Associate Professor, Franklin & Marshall College, Department of Business, Organizations, and Society, Lancaster, PA, 17604-3003, United States, [email protected], Dennis C. Bely
A methodology for constructing cost-effective live-fire test and evaluation (LFT&E) programs within the MMF environment is presented. Issues addressed include the design and execution of LFT&E programs to ensure the collection of data relevant to the assessment of system vulnerability/lethality and the evaluation of test results in a format useful to decision-makers concerned with accomplishing system of systems collective tasks and achieving mission success in the joint environment.

3 - Marks of Distinction: Style as a Source of Status among Security Analysts, 1986-2005 Simona Giorgi, Northwestern University, Kellogg School of Management, 2001 Sheridan Road, Evanston, IL, 60208, United States, [email protected], Klaus Weber
The study examines the performative basis of status in occupational communities, status that derives from the style with which an actor performs a job. We examine status dynamics among securities analysts specializing in biotechnology using text analysis of analysts’ reports. We find that stylistic elements such as the rhetorical packaging of reports are associated with status if they resonate with audience values.

I SB25
Panel Discussion: OR Entrepreneurship #2
Cluster: OR Entrepreneurship Invited Session
Chair: James Kelly, CEO, OptTek Systems Inc., 1919 Seventh Street, Boulder, CO, 80302, United States, [email protected] 1 - Panel Discussion: Entrepreneurship in OR/MS - Lessons Learned Moderator: James Kelly, CEO, OptTek Systems Inc., 1919 Seventh Street, Boulder, CO, 80302, United States, [email protected], Panelists: Dennis Pegden, Ron Laughery, Mark Elder, James Franklin
Many members of INFORMS have worked in an entrepreneurial capacity. Some have started and/or grown companies on a full-time/ or part-time basis. The purpose of this session is to recount some of the stories associated with these activities and to discuss common experiences, obstacles, and strategies. This group has a particular focus on products and services related to simulation and optimization.

I SB23
CONDIE (Committee on Next Decade in Industrial Engineering) Progress Report
Cluster: New Frontiers at the Intersection of Industrial Engineering and O.R. Invited Session
Chair: Lawrence Seiford, Professor and Chair, University of Michigan, 1205 Beal Avenue, Ann Arbor, MI, 48109-2117, United States, [email protected] 1 - Panel: CONDIE Progress Report Moderator: Lawrence Seiford, Professor and Chair, University of Michigan, 1205 Beal Avenue, Ann Arbor, MI, 48109-2117, United States, [email protected]
The Committee on the Next Decade in Industrial Engineering (CONDIE) was formed to help ensure that academia will be able address areas of critical need in both research and education over the coming decade. CONDIE will serve as a continuing forum for applied dialog and deliberate action focused on facilitation of a timely response to national need, while advancing the frontier of engineering research.

75

SB26 I SB26

INFORMS PITTSBURGH — 2006
used to evaluate the historical data for expected column behavior and to decipher signatures in the common cause variability that indicate problems presenting opportunities for process improvement. This session focuses on the engineering issues that must be considered in tailoring the statistical analyses towards that end.

Data Mining and Statistical Modeling in Environmental Problems
Sponsor: INFORMS Computing Society/Data Mining Sponsored Session
Chair: Seoung Bum Kim, Assistant Professor, Industrial & Manufacturing Systems Engineering, The University of Texas at Arlington, PO Box 19017, Arlington, TX, 76019, United States, [email protected] 1 - Space-Time Modeling of Total Column Ozone Levels Mikyoung Jun, Assistant Professor, Department of Statistics / Texas A&M University, College Station, TX, 77843, United States, [email protected]
For space-time processes on global or large scales, it is critical to respect the Earth’s spherical shape. The cov fns of such processes should not only be positive definite, but be capable of capturing the dynamics. We develop space-time cov fns on sphere-time that are flexible for space-time interactions, especially spacetime asymmetries. Our approach yields explicit expressions for cov fns, which has great advantages in computation. We apply our model to global total column ozone levels.

2 - EWMA and CUSUM Control Chart Using Wavelets for Complicated Functional Data Young S. Jeong, Student, University of Tennessee, 309 EST, Knoxville, TN, 37996, United States, [email protected], Myong K. Jeong, Mary Leitnaker
This paper extends the adaptive thresholding to the versions of the exponentially weighted moving average (EWMA) and cumulative sum (CUSUM) control for process monitoring based on functional data. Simulation studies show that proposed methods quickly detect smaller local and vertical shifts.

I SB28
Best Student Paper Presentation Session
Sponsor: Quality, Statistics and Reliability Sponsored Session
Chair: Judy Jin, Associate Professor, Department of Industrial and Operations Engineering, University of Michigan, IOE 2855, Ann Arbor, MI, 48109, United States, [email protected] 1 - Robust Synthesis of Nanostructures Tirthankar Dasgupta, PhD Student, Georgia Institute of Technology, 765 Ferst Drive NW, Atlanta, GA, 30332, United States, [email protected], Christopher Ma, Roshan Vengazhiyil, Zhong Lin Wang, Jeff C. F. Wu
An effort is made to systematically investigate the best process conditions that ensure large-scale synthesis of different types of nanostructures. Models linking the probabilities of obtaining specific morphologies to the process variables are developed. A new iterative algorithm for fitting a Multinomial GLM is proposed and used. The optimum process conditions that are robust with respect to inner noise are derived from the fitted models using Monte-Carlo simulations.

2 - Characterization of Spatial and Temporal Patterns of PM2.5 in the Continental United States Chivalai Temiyasathi, PhD Student, Industrial & Manufacturing Systems Engineering/University of Texas at Arlington, Arlington, TX, 76019, United States, [email protected], Seoung Bum Kim
Statistical analyses for time series or spatial data have been widely used to investigate the behavior of ambient air pollutants. The main objective of this study is to characterize the spatial correlation and the temporal pattern of PM2.5. Clustering analysis with correlation distance was performed to group the monitoring stations where the clustered stations have similar temporal patterns over the monitoring period. Moreover, the temporal patterns in each clustered region were investigated.

3 - The Impact of Peaking Electrical Generating Units in Northeast Alper Unal, MACTEC Inc., Trenton, NJ, United States, [email protected]
Electric Generating Units are one of the highest contributors of air pollutants, especially for NOX and SO2. Under certain conditions their impact on air quality might be significant. In this study, our aim is to quantify uncertainty in EGU emissions. For this we have developed unit specific temporal profiles, using hourly data for Northeastern United States. This paper will present detailed analysis emission processing as well as its implications on air quality through CMAQ model.

2 - Robust Parameter Design for Quality Variables and Reliability Measures Lingyan Ruan, Ph.D. Student, Georgia Institute of Technology, Industrial and Systems Engineering, 765 Ferst Drive, Box 0205, Atlanta, GA, 30332, United States, [email protected], Jye-Chyi (JC) Lu
The literature in design of experiment (DOE) considering both quality and reliability metrics is scarce. Although product reliability is as important as quality, especially for electronic or semiconductor devices, robust parameter designs (RPD) are primarily applied to quality measures. Moreover, typical DOE for accelerated degradation tests (ADTs) does not explore impact of changing manufacturing conditions on product reliability. This article develops DOE plans for manufacturing variables and ADT conditions to minimize variance of parameter estimates and RPD optimizations for selecting controllable manufacturing variables to achieve longest product lifetime and reduce variations induced by environmental noise.

4 - Discovery of Preference Information via Data Mining Techniques in Multi Criteria Decision Making Ozge Kaplan, PhD Candidate, Department of Civil, Construction and Environmental Engineering, North Carolina State University, Raleigh, NC, United States, [email protected], S. Ranji Ranjithan
An iterative method that integrates data mining algorithms with optimization methods is developed to help decision makers (DMs) find the best compromise solution. DMs’ rankings of systematically generated Pareto- and near-Paretooptimal alternatives are processed using data mining algorithms to discover DMs’ implicit preferences, both in objective space and decision space. These rankings are then used in the search method. This method is being applied to a solid waste management problem.

3 - Defect Pattern Recognition in Semiconductor Fabrication using Model-based Clustering and Bayesian Tao Yuan, The University of Tennessee, [email protected], Way Kuo
Spatial defect patterns on semiconductor wafers contain useful information about potential problems in the manufacturing processes. This study proposes to use model-based clustering via Bayesian inference for detecting defect patterns. The new clustering method detects the defect clusters and identifies the pattern of each cluster automatically. Promising results are obtained through simulation studies.

I SB27
Process Mining
Sponsor: Quality, Statistics and Reliability Sponsored Session
Chair: Mary Leitnaker, Professor, University of Tennessee, Department of Statistics, Knoxville, TN, 37996, United States, [email protected] 1 - Statistical Analyses for Process Improvement: A Distillation Case Study Nitin Kaistha, Assistant Professor, Indian Institute of Technology, Department of Chemical Engineering, Kanpur, 208016, India, [email protected], Mary Leitnaker
Distillation is the most ubiquitous of unit operations in the chemical industry. An array of statistical tools ranging from univariate to multivariate methods can be

4 - Design of DOE-based Automatic Process Controller with Consideration of Model Uncertainties Jing Zhong, University of Michigan, [email protected], Jianjun Shi, Jeff Wu
This paper developed a DOE-based automatic process control scheme considering both observation and modeling uncertainties. The simulation demonstrates that considering the uncertainties, automatic controllers can achieve better process performance than conventional off-line design, and more stable than normal APC controllers. The proposed approach is illustrated using an industrial process.

76

INFORMS PITTSBURGH— 2006 I SB29
Healthcare Engineering to Inform Policymakers
Cluster: Joint Cluster Healthcare/ HAS: Healthcare Engineering Invited Session
Chair: Julie Swann, Assistant Professor, Georgia Institute of Technology, Industrial and Systems Engineering, Atlanta, GA, 30332, United States, [email protected] Co-Chair: Daniel Faissol, PhD Student, Georgia Institute of Technology, School of ISyE, Atlanta, GA, 30332, United States, [email protected] 1 - The Benefits of a Time-Varying Objective Function: Harm Reduction Over the Course of a Drug Epidemic Jon Caulkins, Professor of Operations Research and Public Policy, Heinz School of Public Policy and Management, Carnegie Mellon University - Qatar, PO Box 24866, Doha, Qatar, [email protected], Andrea Seidl, Gernot Tragler, Dagmar Wallner
A central debate in drug policy concerns the relative merits of “harm reduction” (e.g., HIV/AIDS control) vs. “use reduction” (controlling use per se). This paper models whether shifting emphasis over the course of a drug epidemic can improve outcomes for heroin in Australia and cocaine in the US. of patients consulted.

SB31

2 - An Analysis of the Pediatric Vaccine Supply Shortage Problem Sheldon Jacobson, Professor, University of Illinois, 201 N. Goodwin Avenue, Urbana, IL, 61801, United States, [email protected], Edward Sewell, Ruben Proano
The impact of pediatric vaccine stockpile levels on immunization coverage rates is analyzed. The results suggest that the CDC/HHS proposed vaccine stockpile levels are adequate for production interruptions that last no more than six months. However, given that recent production interruptions have lasted (on average) for over one year, the proposed stockpile levels are insufficient during such time periods, which in turn could lead to disease outbreaks.

3 - Dynamic Multi-Priority Patient Scheduling with Uncertain Demand Jonathan Patrick, PhD, University of British Columbia, 1215 Duthie Avenue, Burnaby, BC, V5A 2R5, Canada, [email protected], Martin L. Puterman, Maurice Queyranne
We present a method for scheduling multi-priority patients to a diagnostic facility in a public health setting. Rather than maximize revenue, the resource manager seeks to allocate capacity so that waiting time targets are met. Our research involves modeling the scheduling process as an MDP and applying approximate dynamic programming methods. We present the optimal linear approximation for two cost structures as well as the resulting booking policy for achieving the waiting time targets.

2 - Optimizing Pediatric Vaccine Formularies Sheldon Jacobson, Professor, University of Illinois, 201 N. Goodwin Avenue, Urbana, IL, 61801, United States, [email protected], Edward Sewell, Shane Hall
This presentation introduces the Vaccine Formulary Selection with Limited Budget Problem for obtaining the optimal childhood immunization vaccine formulary. The model that is formulated allows public health policy makers and pediatricians to evaluate the combinatorial explosion of vaccine choices spawned by the development of combination vaccines.

I SB31
Topics in Portfolio Optimization
Cluster: Financial Engineering and Risk Management Invited Session
Chair: John R. Birge, Professor, University of Chicago, Graduate School of Business, 5807 S. Woodlawn Avenue, Chicago, IL, 60637, United States, [email protected] 1 - Hedging Multi-Period Guarantees in the Presence of Transaction Costs via Stochastic Programming Stein-Erik Fleten, Associate Professor, NTNU Norway, IØT, Alfred Getz v. 3, Trondheim, NO, NO-7491, Norway, [email protected], Snorre Lindset
Multi-period guarantees are often embedded in life insurance contracts. For the multi-period guarantee there is a rather large rebalancing of the hedge portfolio as we go from one period to the next. By introducing transaction costs, the size of this rebalancing is reduced. Transaction costs may therefore be one possible explanation for why we do not see the insurance companies performing a large rebalancing of their investment portfolio at the end of each year.

3 - Quantifying the Routes of Transmission and Assessing Nonpharmaceutical Interventions for Influenza Lawrence Wein, Paul E. Holden Professor of Management Science, Graduate School of Business, Stanford University, 518 Memorial Way, Stanford, CA, 94305, United States, [email protected], Michael Atkinson
We construct a hierarchical model for pandemic influenza, which embeds a three-compartment model of the household within a household epidemic model. We use data to quantify airborne vs. contact transmission for influenza, and assess the effectiveness of various nonpharmaceutical interventions within the household during a pandemic influenza outbreak.

4 - The Cost Effectiveness of Screening Newborns for Genetic Diseases Using Tandem Mass Spectrometry Greg Zaric, Professor, Ivey School of Business, University of Western Ontario, London, ON, Canada, [email protected], Tony Rupar, Lauren Cipriano
We developed a model to investigate the health effects, costs and cost effectiveness of using tandem mass spectrometry (MS/MS) to screen for “bundles” of up to 21 diseases. We found that it is not cost effective to use MS/MS to screen for a single disease, but there are many bundles of diseases for which the average cost effectiveness is between $50,000 and $100,000 per life year gained. It is not cost effective to screen for all diseases that can be tested for using MS/MS.

2 - The Impact of Consumption and Liquidity Constraints on Optimal Consumption and Investment Decisions Amit Bhandari, Northwestern University, 2145 N Sheridan Rd, Evanston, IL, United States, [email protected], John R. Birge
Both individual and institutional investors face a number of constraints in their consumption and investment decisions. We look at well-motivated constraints on the consumption process as well as liquidity constraints and study their impact on optimal consumption and investment policies under a dynamic discrete time setting.

I SB30
Pierskalla Award Presentations
Sponsor: Health Applications Section Sponsored Session
Chair: Julie Simmons Ivy, Assistant Professor, University of Michigan, 701 Tappan Street, Ann Arbor, MI, 48109, United States, [email protected] 1 - A Mean-Variance Model to Optimize the Fixed vs. Open Appointment Percentages in an Open-Access Clinic Xiuli (Shelly) Qu, Purdue University, 315 N. Grant Street, West Lafayette, IN, 47906, United States, [email protected], Ronald L. Rardin, Julie A. S. Williams
To improve the accessibility to outpatient clinics, open access scheduling provides open appointment slots for short-notice scheduling. To determine the percentage of open appointments, we introduce a mean-variance model, and develop a procedure to search all Pareto optimal solutions to the model. Our numerical results show that provider capacity and average demand for open appointments considerably affect the Pareto optimal expectations and variances of the number

3 - Portfolio Selection Problem with Transaction Costs: An Option Pricing Approach Zhen Liu, PhD Candidate, Northwestern University, 2145 Sheridan Road, Department of IE/MS, Evanston, IL, 60208, United States, [email protected], John R. Birge
We studied the modeling and numerical solution of portfolio selection problem with various transaction costs: In analogous to American option pricing problems, our problem is modeled as buy-and-hold portfolio problems plus optimization ones. Buy-and-hold problems are solved by Local Discontinuous Galerkin method to achieve higher order of accuracy. We incorporate transaction costs in the optimization problems to find the non-trade region. Last convergence of our algorithm will be discussed.

4 - Simulation of Cointegrated Returns for Fund-of-Funds Construction Under Different Risk Measures Alexander Galenko, University of Texas at Austin, 1 University Station, C2200, Austin, TX, 78712, United States, [email protected], Elmira Popova, Ivilina Popova
We construct a new sampling algorithm for cointegrated hedge funds returns and create a portfolio of funds using different risk measures. We compare the performance of this portfolio when the returns are assumed to be correlated.

77

SB32 I SB32
Consumer Level Financial Modeling
Sponsor: Financial Services Sponsored Session

INFORMS PITTSBURGH — 2006
3 - Analysis of Air Transportation System Delay John-Paul Clarke, Associate Professor, Georgia Institute of Technology, School of Aerospace Engineering, 270 Ferst Drive, Atlanta, GA, 30332, United States, [email protected], Terran Melconian
Flight delays are a growing problem in the US air transportation system. However, the complex interactive nature of the system makes it difficult to quantify the systemwide impact of specific delays from historical data. We present the results of a study where the impact of given absolute and relative delays are quantified through simulation for the different phases of operation, thereby enabling a comparison of the likely effectiveness of delay reducing technologies.

Chair: Aparna Gupta, Assistant Professor, Rensselaer Polytechnic Institute, 110 8th Street, Troy, NY, 12180, United States, [email protected] 1 - Credit Risk Modelling for Portfolios of Consumer Loans Lyn Thomas, Professor of Management Science, School of Management, University of Southampton, Southampton, SO17 1 BJ, United Kingdom, [email protected]
There are more models on credit risk in portfolios of corporate loans compared to consumer loans, even though lending in the latter is 70% higher. We describe three approaches to building such models which have some relationship with structural models, reduced form default mode models and reduced form mark to market models in the corporate case even though they differ substantially from such models. The performance of the models using UK consumer lending data is investigated.

4 - Improving the Performance of Monitor Alert: Estimating Delay Distributions for Wheels-up Time Yufeng Tu, Graduate Research Assistant, R. H. Smith School of Business, University of Maryland, College Park, MD, 20742, United States, [email protected], Michael Ball, Wolfgang Jank
In this paper we present an extension of our previous approach for estimating flight departure delay distributions. Our models include componenets for seasonal trend, daily propagation and day-of-the-week effects. We also explore flight auto-correlations as a means toward creating a dynamically reactive model.

2 - A Wealth Risk Management Framework to Achieve Protective and Aspirational Objectives Aparna Gupta, Assistant Professor, Rensselaer Polytechnic Institute, 110 8th Street, Troy, NY, 12180, United States, [email protected], Zhisheng Li
This paper formulates a wealth allocation framework to achieve downside wealth protection as well as an upside potential of affluence. Three risk dimensions are identified from the perspective of individual investors — protective, market and aspirational — and available assets are classified into these based on their riskreturn profiles. All risk dimensions are optimized simultaneously through appropriate allocation of total wealth in the three asset groups.

I SB34
Transportation Infrastructure Management
Sponsor: Transportation Science & Logistics Sponsored Session
Chair: Pablo L. Durango-Cohen, Assistant Professor, Northwstern University, 2145 Sheridan Road, A335, Evanston, IL, 60208-3109, United States, [email protected] 1 - A Dynamic Infrastructure Performance Model for Evaluating Maintenance Effectiveness and Poolability Chih-Yuan Chu, Student, Northwestern University, 2145 Sheridan Road, Evanston, IL, 60208, United States, [email protected], Pablo L. Durango-Cohen
The paper uses state-space specifications of time series and develops dynamic, stochastic, multivariate infrastructure performance models that capture the effectiveness of maintenance activities, exogenous factors and uncertainty. A fully and a partially aggregated specifications are proposed. Their predictive power and poolability of data are evaluated by a benchmark model.

3 - Using Customer Goals to Determine an Appropriately Risky Strategy Sergio Granville, Power Systems Research Institute, [email protected], Mike McCoy
This presentation describes the on-going search for better investment strategies when there is a perceived skew in intermediate market returns. The original implementation was to sell the missing tail of the return distribution. Marketing this idea uncovered the customer concern expressed as “picking up nickels in front of a steam roller”. This reality check lead to a reshaping of the problem. Is there an investment strategy that meets customer goals, with risk tolerable to the customer?

I SB33
Delay Models for Aviation Operations
Sponsor: Aviation Applications Sponsored Session
Chair: David Lovell, Associate Professor, University of Maryland, Department of Civil Engineering, 1173 Martin Hall, College Park, MD, 20742, United States, [email protected] 1 - Identification and Adjustment of Inaccurate Airport Arrival Capacities Andrew Churchill, Graduate Research Assistant, University of Maryland, College Park, MD, 20742, United States, [email protected], David Lovell, Michael Ball, Avijit Mukherjee
Modeling flight delays and cancellations at an airport using statistical and analytical methods require as input the airport capacity during different time intervals of a day. In this paper, we present a statistical model to estimate the arrival capacity of an airport during any time interval, based on several factors such as runway configuration, meteorological conditions, etc. We show that these estimates reduce the prediction error of existing models for delays and cancellation.

2 - Determining Economic Maintenance Schedule for Interdependent Components Pattharin Sarutipand, Student, Northwestern University, 2145 Sheridan Road, Evanston, IL, 60208, United States, [email protected], Pablo L. Durango-Cohen
Optimal group maintenance policies for a system of functionally interdependent components are considered. We formulate the problem as a constrained nonlinear mixed-integer programming. A heuristic method is proposed for coordinating maintenance schedules for the clusters of components. An example is given to illustrate the method.

3 - An Analytical Solution for the Finite-Horizon Pavement Resurfacing Planning Problem Yanfeng Ouyang, Assistant Professor, University of Illinois at Urbana-Champaign, 205 N. Mathews Avenue, Urbana, IL, 61801, United States, [email protected], Samer Madanat
This paper presents an analytical approach for planning highway pavement resurfacing activities in the case of continuous pavement state and continuous time. It derives analytical optimality conditions and develops a simple algorithm to solve for the optimal resurfacing frequency and intensity that minimize lifecycle costs in a finite horizon.

2 - Relationship Between En Route Delay and System Delay in the NAS Mark Hansen, Professor, University of California at Berkeley, Civil and Environmental Engineering, 114 McLaughlin Hall, Berkeley, CA, 94720, United States, [email protected], Helen Yin
As air traffic increases, en route bottlenecks are becoming increasingly important factors in NAS operating performance. In cases where airports are also bottlenecks, the system effect of en route bottlenecks and the delays resulting from them depend on the details of their timing in relation to queuing conditions at the destination airport. We present a simple method for measuring the system effect of an en route delay and results of measurements using this method.

4 - Un-Staffed Facilities in the National Airspace System: Current Status and Future Work Jasenka Rakas, University of California at Berkeley, NEXTOR, 107B McLaughlin Hall, Berkeley, CA, 94720, United States, [email protected], Gautam Gupta
This study develops a sampling methodology to comprehensively evaluate over 9000 unstaffed facilities of the National Airspace System infrastructure assets. Because the FAA has near accurate information for some of the facilities with very high repair costs, we define a sampling methodology for the rest of the facilities based on such incomplete data. We approach the problem as an exercise in stratified sampling, with each strata representing certain set of facility over a region.

78

INFORMS PITTSBURGH— 2006 I SB35
Models and Methods for Road Pricing: I
Sponsor: Transportation Science & Logistics Sponsored Session
Chair: Yafeng Yin, Assistant Professor, University of Florida, 365 Weil Hall, Box 116580, Gainesville, FL, 32611, United States, [email protected] 1 - Dynamic Tolling Strategies for Managed Lanes Yingyan Lou, University of Florida, Department of Civil & Costal Engineering, University of Florida, Gainesville, FL, 32611, United States, [email protected], Yafeng Yin
This paper proposes two practical but sensible approaches to determine pricing strategies for operating managed or express toll lanes. The tolls vary dynamically in response to real-time traffic conditions in order to provide a superior free-flow travel service to the users of the toll lanes as well as maximize the freeway’s throughput rate. A simulation study is conducted to validate and compare the proposed approaches.

SB37

2 - Toward Engineering Design of Factories Leon McGinnis, Georgia Institute ofTechnology, 765 Ferst Drive, Atlanta, GA, 30332, United States, [email protected], Edward Huang, Kan Wu, Diego Tamburini
Engineers who design products routinely use powerful computer based and integrated design and analysis tools. Engineers who design factories routinely use stand alone tools like Excel, AutoCAD, and simulation, and rarely—if ever—use the results of research on factory design because it isn’t an option in the tools they use. This talk describes how we can change that, and give factory designers access to the same kinds of powerful, integrated tools that product designers use.

3 - A New Mixed Integer Programming Formulation for Facility Layout Design Using Flexible Bays Sadan Kulturel-Konak, Assistant Professor of MIS, Penn State University - Berks, Tulpehocken Rd. PO Box 7009, Reading, PA, 19610, United States, [email protected], Abdullah Konak, Alice Smith, Bryan A. Norman
A mixed-integer programming formulation is presented to find optimal solutions for the block layout problem with unequal departmental areas arranged in flexible bays. The nonlinear department area constraints are modeled in a continuous plane without using any surrogate constraints. The formulation is extensively tested on problems from the literature.

2 - A Day-to-Day Dynamic Congestion Pricing Policy Towards System Optimal Fan Yang, ESRI, Inc., 380 New York Street, Redlands, CA, 92373, United States, [email protected]
This study proposed a simple and socially efficient day-to-day dynamic pricing policy for transport agencies to guide the day-to-day dynamic traffic flow movement to system optimal instead of user equilibrium, considering a more general adjustment behavior and less restrictive assumptions on the link travel time function.

4 - Facility Relayout to Alleviate Congestion Rakesh Nagi, Professor, Department of Industrial & Systems Engineering, 438 Bell Hall, University at Buffalo (SUNY), Buffalo, NY, 14260, United States, [email protected], Rajan Batta, Min Zhang
As product mix and volumes change, certain areas of the manufacturing facility become congested. Some of this congestion can be alleviated by rerouting flows. However, a manager needs to know when to relayout the facility to not only have reduced material handling (traditional objective function) but also reduced congestion. We present facility layout models that explicitly consider flow congestion.

3 - Agent-Based Modeling of Network Demand, Pricing, and Investment Dynamics Lei Zhang, Assistant Professor, Civil, Construction, and Environmental Engineering, Oregon State University, 220 Owen Hall, Corvallis, OR, 97331, United States, [email protected]
The welfare effects of road pricing depend on the decisions by users, owners, and regulators. Built on a novel travel demand and assignment model where each traveler is capable of learning and communicating, and has unique characteristics (e.g. value of time), this paper adopts consistent agent-based techniques to model the decision-making processes of those players and their interactions to address economic and policy analysis needs.

I SB37
Business Value of Information Technology
Sponsor: Information Systems Society Sponsored Session
Chair: Indranil Bardhan, Assistant Professor, The University of Texas at Dallas, School of Management, SM 41, 2601 N. Floyd Road, Richardson, TX, 75083-0688, United States, [email protected] 1 - Managing Value and Uncertainty in Service-Driven IT Investments Ryan Sougstad, Doctoral Student, University of Minnesota, Carlson School of Management, Minneapolis, MN, 55455, United States, [email protected], Robert Kauffman
Service-driven IT solutions require a shift in IT investment evaluation from methods that assess fixed costs of acquisition toward methods that emphasize variable cost-based service solutions for which contract structure is key. Contracts must balance flexibility and quality with profit and risk tolerance. We apply a technique called value-at-risk to evaluate the profitability of IT service contracts, along with the impact a portfolio of contracts will have on the firm’s overall IT risk profile.

4 - Area-Based Pricing Siriphong Lawphongpanich, Associate Professor, Industrial and Systems Engineering, University of Florida, PO Box 11-6595, Gainesville, FL, 32611, United States, [email protected], Yafeng Yin
In this talk, we consider the problem of finding a one-time charge for using roads in a given area. The objective is to reduce total travel time and the area need not be contiguous or form a cut-set. However, when the area forms a cut-set and there are additional constraints on tollable roads, then the problem reduces to the cordon pricing problem in the literature.

I SB36
Manufacturing Facility Design
Sponsor: Transportation Science & Logistics Sponsored Session
Chair: Rakesh Nagi, Professor, Department of Industrial & Systems Engineering, 438 Bell Hall, University at Buffalo (SUNY), Buffalo, NY, 14260, United States, [email protected] 1 - Redesign of Retail Store Layouts Bryan A. Norman, PhD, University of Pittsburgh, 1033 Benedum Hall, Pittsburgh, PA, 15261, United States, [email protected], Alice Smith, Chen Li
This talk considers detailed layout design and optimization for retail stores. Given initial aisle settings and preliminary department assignments, we redesign the aisle structure including constructing additional aisles, changing aisle widths, and locating the departments in each sub-area of the facility with adjusted sizes to maximize the overall profit. We utilize a concave linear piecewise profit per unit area function rather than a constant one for all departments in the retail area.

2 - Does Offshoring Impact Customer Satisfaction? Jonathan Whitaker, Doctoral Candidate, University of Michigan, Ross School of Business, Ann Arbor, MI, United States, [email protected], M.S. Krishnan, Claes Fornell
While researchers have noted an increase in offshoring, there is limited work on the implications for firms and consumers. This research begins to address these implications by studying the relationship between firms’ offshoring activities and their American Customer Satisfaction IndexTM scores. We analyze panel data of 68 firms from 1997-2004, and find that offshoring is associated with an increase in perceived value and perceived quality, the two key components of customer satisfaction.

3 - Effects of Information Transparency on Supply Chain Coordination Under Competition Kevin Zhu, Associate Professor, University of California, The Rady School of Management, San Diego, CA, 92093-0093, United States, [email protected], Zach Zhou
Information sharing has been shown to benefit a supply chain. Developing a game theoretic model, we show that information transparency can create value, yet it affects retailers and manufacturers very differently: one side will always be hurt. This means that the theoretical benefits of information sharing will not materialize, a fundamental challenge with information sharing in supply chains. To deal with this issue, a discriminatory pricing scheme is proposed to internalize the incentives.

79

SB38

INFORMS PITTSBURGH — 2006 I SB39
Where Do We Want to Go in Research?
Cluster: Where Do We Want to Go?
(In observance of Arthur Geoffrion’s retirement)

4 - Adoption of New Trading Technology: Firm-Specific Effects and Network Effects in Options Exchanges Bruce Weber, Professor, London Business School, Regents Park, London, NW1 4SA, United Kingdom, [email protected]
Diffusions of IT such as email, browsers, and electronic markets are sensitive to feedback. Realized impacts arise from technical progress and economic forces (adoption and installed base). The all-electronic International Securities Exchange and Boston Options Exchange, opened in 2000 & 2004, and compete with 4 US floor markets. Tobit adoption models estimated from 350 quarterly disclosures from 20 major brokers show firm-specific factors explain 60% of e-exchange use and network effects 40%.

Invited Session
Chair: Charles Corbett, Associate Professor, UCLA Anderson School of Management, 110 Westwood Plaza, Los Angeles, CA, 90077, United States, [email protected] 1 - Leading a Double Life Uday Karmarkar, Professor, UCLA Anderson, 110 Westwood Plaza, Box 951481, Los Angeles, CA, 90077, United States, [email protected]
The difficulties of balancing research with education and practice are well known. There are a number of ways of achieving your personal goals with respect to those three areas, but most involve doing double duty, and working quite hard. Still the rewards and satisfactions are also substantial.

I SB38
Railway Decision Support Models
Sponsor: Railroad Applications Sponsored Session
Chair: David Hunt, Cambridge Systematics, Inc., PO Box 816, Pennington, NJ, 08534-0816, United States, [email protected] 1 - A Web-Based Decision Support System for Railroad Service Design Ravindra Ahuja, Professor, University of Florida, 303 Weil Hall, Gainesville, FL, 32611, United States, [email protected]
Railroad service design, which comprises developing a railroad’s operating plan, has been one of the most challenging planning problems faced by US freight railroads and has been solved manually so far. We will describe and demonstrate a web-based interactive decision support system that enables a railroad to determine its operating plan within a week and incrementally change its current operating plan within days.

2 - It is Not a Fair World! Don Ratliff, Regents and UPS Professor, Georgia Institute of Technology, School of ISyE, Atlanta, GA, 30332, United States, [email protected]
If you are really smart, really want to impact practice, work really hard, and are really prolific at publishing in our great journals, it seems reasonable to expect industry to beat a path to your door. Since this “beating a path to your door” has happened to only a few researchers in our field: are these researchers smarter, do they work harder, is it just dumb luck, is there some magic formula, or is the world just not fair? This session will provide the definitive answer.

2 - A Decision Support System for Hump Yard Management Larry Shughart, Vice President, Business Development, Innovative Scheduling, Inc., 2153 SE Hawthorne Road, Suite 128, Gainesville, FL, 32641, United States, [email protected], Ravindra Ahuja, Nikhil Dang, Arvind Kumar, Saurabh Mehta
Efficient hump yard operations are critical for any freight railroad. By integrating simulation with optimization, we have developed a decision support system for hump yard management that can be used for planning, strategy as well as realtime scheduling. For example, it can be used to determine the blocking and car handling capacities of a yard and the impact of yard resources (crew and locomotives) on yard performance.

3 - Greatest Hits in Operations Research Gerald Brown, Professor, Operations Research Department, Naval Postgraduate School, Monterey, CA, 93943, United States, [email protected]
What makes a discovery in operations research so fundamental it demands mastery from any compleat analyst? Consider your personal collection of OR’s greatest hits (you maintain one, don’t you?). Regardless of medium- refereed publication, lecture, book, computer program, course handout, press interview what motivates your selections? The list I recommend to my colleagues and students includes some obscure entries, and fewer winners of our well-known awards than you might suspect. Why is this? What are the distinguishing telltales of a fundamental contribution, and how do we use these to discover as early as possible the importance of such a signal result?

3 - Algorithms for the Train Platforming Problem (TPP) Laura Galli, PhD Student, DEIS - Universitá degli Studi di Bologna, Viale Risorgimento 2, Bologna, 40136, Italy, [email protected], Alberto Caprara, Michele Monaci, Paolo Toth
We present an ILP model and an exact algorithm for a specific type of TPP arising in real life applications. The goal is to assign each train a platform and two paths, avoiding platform-conflicts, keeping the number of path-conflicts under a given threshold and maximizing the assignment quality. The formulation has a large number of variables and constraints, thus we use a Branch-and-Price approach and a separation procedure requiring the solution of a specialized Maximum Weight Clique Problem.

4 - Models and Managers: Then, Now, and in the Future John Little, Professor, MIT Sloan School, Room E56-308, 38 Memorial Drive, Cambridge, MA, 02142, United States, [email protected]
Decision support systems have evolved rapidly due to the explosion of available data, increased computer power, and advances in modeling methods. I shall review highlights of the evolution of DSSs from the point of view of managerial models, focusing on marketing, and offer a few thoughts about what may lie ahead.

I SB40
New Location Models and Algorithms
Sponsor: Location Analysis Sponsored Session
Chair: Lezhou Roger Zhan, Industrial and Systems Engineering, Univeristy of Florida, 303 Weil Hall, Gainesville, FL, 32611, United States, [email protected] 1 - Numerical and Simulation Studies on a Supply Chain Design Problem with Supply Disruptions Lian Qi, University of Florida, 303 Weil Hall, PO Box 116595, Gainesville, FL, 32611-6595, United States, [email protected], Zuo-Jun Max Shen
We study an integrated supply chain design problem that includes one supplier, one or more retailers and customers. Random disruptions at both the supplier and retailers are considered in our research. We conduct extensive numerical and simulation experiments to obtain managerial insights into this supply chain design problem.

4 - Solving a Real-World Train Unit Assignment Problem Valentina Cacchiani, PhD Student, Universita di Bologna, Viale Risorgimento 2, 40136 Bologna, Italy, [email protected], Alberto Caprara, Paolo Toth
We study the problem of creating rosters for a set of traction units for a railway transportation company. The aim is to use the smallest number of traction units, so as to cover a planned timetable of trips, by imposing a minimum number of places required for each trip and other technical constraints. We propose integer linear programming models and heuristics based on the associated linear programming relaxations. Computational results on real-world instances are presented.

80

INFORMS PITTSBURGH— 2006
2 - The Effects of Customer Behavior On Closed Loop Supply Chain Network Design Kristin Sahyouni, Northwestern University, 2145 Sheridan Road, Department of IEMS, Evanston, IL, 60208, United States, [email protected], Mark Daskin, Canan Savaskan
We investigate how the level of customer participation in a voluntary returns collection program is influenced by the ease of the return process and financial incentives provided (or financial penalties imposed). We subsequently examine how a firm’s network design decisions and various returns acceptance policies are impacted by customer behavior in a CLSC environment. Results and insights are discussed.

SB43

I SB42
Contemporary Scheduling
Cluster: Scheduling Invited Session
Chair: Joseph Leung, Distinguished Professor, New Jersey Institute of Computer Science, Department of Computer Science, Newark, NJ, 07102, United States, [email protected] 1 - Online Scheduling with Known Arrival Times Nicholas G. Hall, Ohio State University, 2100 Neil Avenue, Columbus, OH, United States, [email protected], Marc Posner, Chris Potts
We consider minimizing total weighted completion time in an online scheduling environment where jobs only arrive at known times. We describe an online scheduling algorithm, and show that its competitive ratio is best possible. This result is established by constructing a system of inequalities that characterizes all possible job completion times in a schedule delivered by our algorithm. Two special cases are also studied.

3 - Solution Strategies for Flexible Discrete Location Models Stefan Nickel, Chair of Operations Research and Logistics, Saarland University, Geb A 5 3, Saarland, SL, 66041, Germany, [email protected], Sebastian Velten, Alfredo Marin, Justo Puerto
Ordered Median problems deal, in contrast to classical location problems, with objective functions which are pointwise defined. Nearly all classical objective functions from location theory have an ordered median representation. Moreover, new objective functions can be modeled. We give a new, mixedinteger formulation of the discrete Ordered Median Problem (DOMP). A Branchand-Cut scheme will be presented and computational results will be reported. Moreover, extensions will be discussed.

2 - A Frequency Planning Problem for a Large-Scale Bus Network Jian Yang, Assistant Professor, New Jersey Institute of Technology, Department of Industrial Engineering, Newark, NJ, 07102, United States, [email protected], Steve Chien
For a large-scale bus network with known customer O-D demand levels, we make simultaneous decisions on the opening/closure of locations and link-based bus frequencies with the goal of maximizing the operating profit. We formulate the problem as a mixed integer linear programming (MIP) problem, and rely on the commercial solver LINGO as well as a post-processing heuristic to solve the problem.

4 - Reliability Models for Facility Location: Formulations and Algorithms Lezhou Roger Zhan, Industrial and Systems Engineering, Univeristy of Florida, 303 Weil Hall, Gainesville, FL, 32611, United States, [email protected], Zuo-Jun Max Shen, Jiawei Zhang
In this paper we analyze the facility location problem under the condition that some facilities are not completely reliable. That is, even if a facility is constructed, it is still subject to fail, which will force its customers to get service from a farther operational facility or subject to a penalty cost. We propose approximation algorithms for a simple model where the failure probability at each facility is the same. Some preliminary results about a more general model are also discussed.

3 - An Efficient Optimal Solution to the Two-Yard-Crane Scheduling Problem Weihua Zhou, Hong Kong University of Science & Technology, Clear Water Bay, Kowloon, Hong Kong, [email protected], Chung-Yee Lee
We study a two-yard-crane scheduling problem in a yard block of a port. The two cranes can move along a lane but must maintain a safe distance between them. All containers are ready at the beginning of the period and waiting to be moved from one end of the block to another. We develop efficient algorithms to minimize the maximum tardiness and the number of tardy jobs, respectively.

I SB41
Call Centers
Sponsor: Applied Probability Sponsored Session
Chair: Shane Henderson, Associate Professor, Cornell University, School of ORIE, 230 Rhodes Hall, Ithaca, NY, 14853, United States, [email protected] 1 - Routing and Staffing in Service Systems with Heterogeneous Servers and Impatient Customers Mor Armony, Assistant Professor of Operations Management, Leonard N. Stern School of Business, New York University, 44 West 4th Street #8-62, New York, NY, 10012, United States, [email protected], Avi Mandelbaum
Motivated by call-centers, we consider queueing systems with homogeneous customers and multiple server pools. Servers differ with respect to their speed of service. Customers are impatient and may leave the system before their service starts. For such systems we propose simple staffing and routing rules that are jointly asymptotically optimal in the sense that they minimize staffing costs subject to a constraint on the fraction of abandonment, in the limit, as the arrival rate grows large.

4 - Cooperative Outsourcing Games with Significant Changeover Times Tolga Aydinliyim, Case Western Reserve University, Department of Operations, Cleveland, OH, 44106-7235, United States, [email protected], George Vairaktarakis
We present a cooperative outsourcing model where manufacturers outsource their operations to a single third-party with limited production capacity represented by manufacturing windows. After all manufacturers finish their bookings the third-party offers a new schedule which minimizes the total outsourcing costs of all manufacturers. Coordinated savings are allocated to manufacturers according to a fair allocation rule.

I SB43
New Developments and Applications of the AHP and ANP
Cluster: Analytic Hierarchy Process Invited Session
Chair: Claudio Garuti, General Manager, FULCRUM Ingenieria, Luis Thayer Ojeda 0180 Of.1004, Santiago, RM, Chile, [email protected] 1 - Conflicts Resolution as a Game with Priorities: Multidimensional Cardinal Payoffs Thomas Saaty, University Professor, University of Pittsburgh, 322 Mervis Hall, Pittsburgh, PA, 15260, United States, [email protected]
Two ways to consider increase the effectiveness of game theory in applications 1) By deriving priorities for the payoffs using a cardinal absolute instead of an ordinal or interval scale to do equilibrium analysis; 2) By considering how complex influences are interdependent in non-cooperative situations to derive priorities for best outcomes.

2 - Adaptive Routing in Multi-Skill Call Centers Ger Koole, Professor, Vrije Universiteit, Department of Mathematics, De Boelelaan 1081a, 1081 HV Amsterdam, Netherlands, [email protected], Auke Pot
We introduce an adaptive routing call policy for multi-skill call centers. The objective is to satisfy the service level constraints of all job types as much as possible. This policy does not need prior information on parameter values and allows even for parameter values that change over time. We give ample numerical evidence on the effectiveness of our method.

3 - The Error in Steady-State Approximations for Time-Dependent Performance Measures Shane Henderson, Associate Professor, Cornell University, School of ORIE, 230 Rhodes Hall, Ithaca, NY, 14853, United States, [email protected], Samuel Steckley
Arrival processes to queues invariably exhibit time dependence, but such queues are difficult to analyze. Usually one adopts some kind of steady-state approximation for time-dependent performance. We develop approximations for the resulting error for a range of queueing models. The results reinforce and extend what is known about such errors.

81

SB44

INFORMS PITTSBURGH — 2006
4 - Multi-Player Blackjack Simulation Using a Six-Deck Shoe Donna Retzlaff-Roberts, Professor, University of South Alabama, Mitchell College of Business, Department of Management, Mobile, AL, 36688, United States, [email protected], Alan Chow
Previous blackjack simulations have primarily involved a single deck and a single player. None have involved the six-deck shoe that has become the current standard, and none have investigated the multi-player effect. We simulate blackjack using a six-deck shoe and multiple players to examine the table dynamics that occur in actual blackjack play.

2 - Why We Need AHP/ANP in this World Eizo Kinoshita, Dean/Professor, Meijo University, 4-3-3,Nijigaoka, Kani, 509-0261, Japan, [email protected]
This paper treats human decision making from the perspective of rationality, and defines a Utility Function and Utility Theory as an instrumentally rational decision making theory and AHP as a procedurally rational decision making theory. We show that in practice it is more effective to use the AHP. In this paper we present a partial interpretation of a lasting debate on the effectiveness of Utility Function and the AHP, a debate that does not seem to be ending anytime soon.

3 - Analytical Network Process (ANP) Applications in Turkey Birsen Karpak, Professor of Management, Youngstown State University, One University Plaza, WCBA 635, Youngstown, OH, 44555, United States, [email protected]
The author has implemented Analytical Network Process (ANP) in a variety of her research projects. Two of them will be discussed. An ANP model was developed to identify the priorities of the success factors for Small and Mediumsized Industries (SMEs). An ANP based network to is used to measure the degree of readiness of Turkish industry for total quality management.

5 - Verification, Validation and Quantification of Margins and Uncertainties for Electrical Systems Monica Martinez-Canales, Principal Member of the Technical Staff, Sandia National Laboratories, PO Box 969, MS 9159, Livermore, CA, 94550, United States, [email protected], Genetha Gray
Increasing costs and decreasing resources is encouraging design programs to leverage smaller databases, numerical modeling and simulation activities to support qualification/accreditation assessments. To that end, verification and validation (V&V) and quantification of margins and uncertainties (QMU) must be used to determine simulation-based confidence, predictive capabilities and margin assessments. We will describe aspects of Sandia’s V&V and QMU processes on an electrical circuit model.

4 - Making Psychological Evaluation with AHP/ANP Support Claudio Garuti, General Manager, FULCRUM Ingenieria, Luis Thayer Ojeda 0180 Of.1004, Santiago, RM, Chile, [email protected]
We show a support tool for decision-making in the psychological assessment area, where the huge numbers of variables and knowledge to be structured, integrated and synthesized, forces the need for a system analysis process, able to deal with such complexity. The psychological evaluation system built, called “People Assessment” (PA), captures the following basic pillars: Personality, Competence, and Socio-Economical condition.

I SB45
Tutorial: Math Programming Approaches to Approximate Dynamic Programming
Cluster: Tutorials Invited Session
1 - Math Programming Approaches to Approximate Dynamic Programming Dan Adelman, Professor of Operations Management, University of Chicago, Graduate School of Business, 5807 South Woodlawn Avenue, Chicago, IL, 60637, United States, [email protected]
Approximate Dynamic Programming (ADP) is a new class of methods for breaking Bellman’s curse of dimensionality. We summarize math programming approaches to ADP, focusing on applications in inventory control, revenue management and queueing. We derive a parsimonious, exactly solvable math program whose optimal dual prices are used to construct a value function approximation. Dual information provides economic insights, including a “pricedirected” control policy.

I SB44
Applied Probability II
Contributed Session
Chair: Monica Martinez-Canales, Principal Member of the Technical Staff, Sandia National Laboratories, PO Box 969, MS 9159, Livermore, CA, 94550, United States, [email protected] 1 - Allocation of Processing Time in Multi-Channel Load Balancing Systems Muhammad El-Taha, Professor, University of Southern Maine, 96 Falmouth Street, Deptartment of Mathematics and Statistic, Portland, ME, 04104, United States, [email protected], Bacel Maddah
Consider a new scheme for allocating processing time in a multi-channel loadbalancing system. Servers are divided into two stations where all customers join stations 1 for a fixed amount of service and if necessary join the station 2. Our system outperforms the parallel multi-server model by reducing the mean delay in heavy traffic. Applications will be discussed.

I SB46
Tutorial: A Practical Approach to Pricing Optimization
Cluster: Tutorials Invited Session
1 - A Practical Approach to Pricing Optimization Maarten Oosten, Senior Director of Science and Research, PROS Revenue Management, 3100 Main Street, Suite 900, Houston, TX, 77002, [email protected], Darius Walczak
Pricing optimization generalizes traditional revenue management in that demand need not be independent across products. We define pricing optimization for multiple products facing stochastic demand over a finite time horizon. We address static controls (e.g. LP) as well as dynamic controls. We demonstrate equivalencies between general pricing formulations and traditional revenue management models, and highlight economic and managerial insights. Finally, we address some practical issues, and highlight challenges and open research problems.

2 - Approximate Mean Value Analysis for Closed Queueing Networks with Multiple-Server Stations Rajan Suri, Professor, ISyE Department, University of Wisconsin Madison, 1513 University Avenue, Madison, WI, 53706, United States, [email protected], Sushanta Sahu, Mary Vernon
Closed Queueing Networks are used in manufacturing modeling e.g. for Flexible Manufacturing Systems (FMS) and CONWIP material control. Mean Value Analysis (MVA) is often used to compute the performance measures for these models. The Schweitzer-Bard (S-B) approximation improves the computational efficiency of MVA for single-server stations. We provide a simple extension to SB MVA to enable the analysis of multiple-server stations. Comparisons with simulation show the accuracy of our approach.

3 - Incorporating Time Phased Releases into Open Queueing Networks Diederik Claerhout, Research Assistant, University of Antwerp, Prinsstraat 13, Antwerp, 2000, Belgium, [email protected], Nico Vandaele
One of the tools to control the workload in a system is fine-tuning releases. In contrast with load limited release mechanisms, we did not find satisfying queueing models for time phased release mechanisms. The latter mechanisms can be modelled as imposing minimum cycle times at nodes without blocking servers. Currently, we are studying the impact on the departure processes. We present exact analysis for elementary single server queues and approximating numerical results for the G/G/1 queue.

82

INFORMS PITTSBURGH— 2006 I SB47
Risk, Financing, and Operations
Cluster: Supply Chain and Operations Engineering Invited Session
Chair: Volodymyr Babich, Assistant Professor, University of Michigan, 1205 Beal Avenue, IOE 2783, Ann Arbor, MI, 48105, United States, [email protected] 1 - Operational Hedging Strategies and Competitive Exposure to Exchange Rates Panos Kouvelis, Professor, Washington University in St. Louis, Olin School of Business, One Brookings Drive, Campus Box 1133, St. Louis, MO, 63130-4899, United States, [email protected], Ping Su, Lingxiu Dong
We investigate the impact of operational flexibility on firms’ economic exposure to currency fluctuations in the presence of global competition. We consider a global firm who sells as a monopolist in the domestic market, and also sells to a foreign market facing competition there from a local incumbent. We compare the effects of three operational strategies of the global firm: with no operational flexibility, with postponement flexibility, and with both postponement and allocation flexibility.

SB49

3 - Product-Line Competition: Customization vs. Proliferation Ali Parlakturk, University of North Carolina, Kenan-Flagler Business School, McColl Building, Campus Box 3490, Chapel Hill, NC, 277599, United States, [email protected], Haim Mendelson
We study a market with customers that have heterogeneous preferences for product attributes. We consider two types of firms that compete on price and product variety: A traditional firm, which chooses a limited variety, and a customizing firm, which can produce any configuration. Unlike the traditional firm, the customizing firm does not carry inventory, but its customers incur waiting costs.

4 - Impact of Operational Flexibility on Aircraft Conversion at the US Coast Guard Vinayak Deshpande, Associate Professor, Krannert School of Management, Purdue University, West Lafayette, IN, 47907, United States, [email protected], Asima Mishra, Ananth Iyer
This paper presents the impact of operational flexibility on aircraft conversion from an older to a newer version at the US Coast Guard (USCG). We consider three levels of flexibility: full flexibility, partial flexibility and no flexibility. Both optimization models and Markov Chain based models have been developed to compare aircraft performance under the above three levels of flexibility.

2 - Signaling Value Of Trade Credit Pierre-Yves Brunet, University of Michigan, 1205 Beal Avenue, Ann Arbor, MI, 48109, United States, [email protected], Volodymyr Babich
Start-up firms, denied bank loans because banks are uncertain about creditworthiness of the firms, benefit from trade credit financing because it allows them to implement their production decisions in the current period and because it signals their credit quality to the banks. Using a multi-period model for the operations of the firm, we study joint procurement and financing decisions, characterize properties of the optimal policy, and compute the signaling value of trade credit financing.

I SB49
History of OR/MS II
Cluster: History of OR and Management Science Invited Session
Chair: Heiner Müller-Merbach, Professor, Technische Universität Kaiserslautern, Erwin-Schrödinger-Strafle, Geb. 42-434, Kaiserlautern, D-67653, Germany, [email protected] 1 - German Pre-OR/MS Contributions to OR/MS Heiner Müller-Merbach, Professor, Technische Universität Kaiserslautern, Erwin-Schrödinger-Strafle, Geb. 42-434, Kaiserlautern, D-67653, Germany, [email protected]
Prior to OR/MS, some contributions to the field came from German Business Administration (GBA), such as: (i) Schmalenbach (1873-1955) calculated “shadow prices” for simple pre-LP cases. (ii) Gutenberg (1897-1984) studied preLP bottleneck situations and considered the influence of objectives for production scheduling problems. (iii) Grochla (1921-1986) investigated pre-JIT synchronisation of purchase logistics and production. These and more cases will be presented.

3 - Coordination of Operations and Finance in a Growing Firm Danko Turcic, WSOM, Case Western Reserve University, Cleveland, OH, 44106, United States, [email protected], Matthew Sobel
We address the coordination of operations and finance in a growing firm by optimizing the expected present value of the time stream of dividends in a dynamic model in which demands comprise an ARMA process. The recurring decisions are production quantity, dividends, and short-term borrowing. We study the asymptotic properties of the optimal base-stock level policy for production-inventory. This yields insights into optimal borrowing and production planning for growing firms.

I SB48
Supply Chain Management - Contexts and Concepts
Sponsor: Manufacturing & Service Operations Management Sponsored Session
Chair: Ananth Iyer, Krannert School of Management, Purdue University, West Lafayette, IN, United States, [email protected] 1 - Optimal Timing of Inventory Decisions Using Postponement Options Vishal Gaur, Assistant Professor, Leonard N. Stern School of Business, New York University, 44 West 4th Street, 8-72, New York, NY, 10012, United States, [email protected], Sridhar Seshadri, Marti Subrahmanyam
We consider a single-period inventory model when forecasts of demand and price improve with time. We provide conditions under which postponement is optimal and those for which early exercise takes place. We also determine the impact of risk-premium and volatilities of price-demand on the optimal timing and stocking decisions.

2 - Charles S. ReVelle Richard Church, Professor, Department of Geography, University of California - Santa Barbara, Santa Barbara, CA, 93106, United States, [email protected], John Current
Charles (Chuck) ReVelle passed away last year at the age of 67 from a rare form of cancer. Chuck was a major figure in the field of INFORMS as is evidenced by being the first recipient (with Richard Francis) of INFORMS’s College of Location Analysis award for lifetime achievement. During his career, Chuck wrote 9 books and nearly 200 papers and monographs spanning the fields of OR/MS, Civil Engineering, Economics, Environmental Sciences, Geography; Regional Science, and Water Resource Systems. He had many co-authors (including us), but his favorite and most important was his wife, Penelope.

3 - Fleet Composition of Commercial Jet Aircraft 1952-2005: Developments in Uniformity and Scale Jani Kilpi, Manager, ERP Support, Finnair, Hernemäentie 16, Tuusula, 04330, Finland, [email protected], Ari P.J. Vepsäläinen
The fleet composition of an airline is an important cost driver affecting its operational performance. The fleet composition is characterised here using a fleet uniformity index and a structured way of measuring fleet scale. The history of all jet aircraft operated by commercial passenger or cargo airlines world-wide is analysed for this study, with the results showing that the uniformity of the airline fleets has been steadily decreasing, while their scale has been steadily increasing.

2 - When the Competitor’s Order Quantities are Revealed Who Benefits and Who Loses? Daniela Burkhardt, WHU - Otto Beisheim Graduate School of Management, Burgplatz 2, Vallendar, 56179, Germany, [email protected], Arnd Huchzermeier
On receiving information on competitor’s order quantities, retailers could value the information by adapting both their price and inventory decisions. We explore the benefits of this information for retailer profits and consumer welfare theoretically in a Bertrand model and empirically with scanner data of five major German retailers.

83

SB50 I SB50
Network Design
Sponsor: Telecommunications Sponsored Session

INFORMS PITTSBURGH — 2006
Given a series of past available resources as inputs and the corresponding outputs, we seek the LP model that best explains the outputs. That is, we identify the constraint matrix and objective function of an LP model that fits the data. The maximum decisional efficiency (MDE) estimation principle is used with genetic algorithm search. This is called the Linear Programming System Identification (LPSI) method. Some comparisons with DEA are discussed.

Chair: S. Raghavan, Associate Professor of Management Science, Robert H. Smith School of Business, University of Maryland, 4352 Van Munching Hall, College Park, MD, 20742, United States, [email protected] 1 - Formulations and Reformulations for Network Design Problems with Reload Costs Luis Gouveia, Faculdade de Ciencias da Universidade de Lisboa, DEIO-CIO, Bloco c/2 - Campo Grande, Lisboa, Portugal, [email protected], Ioannis Gamvros, S. Raghavan
We present the notion of reload costs that can appear in several network design and planning problems in telecommunication systems and the transportation industry. We discuss the challenges in modeling such costs with standard approaches and present several reformulations and comment on their strengths and weaknesses.

3 - Measuring Multidimensional Performance Attributes Padma Sastry, Project/Program Manager, The Ohio State University, Page Hall, 1810 College Road, Columbus, OH, 43210, United States, [email protected]
Quality is defined differently depending on the stakeholder: firm, customer or regulator. We develop a method based on DEA applied to 1994-2001 data for the local telephone industry. We analyze trends over time by defining three parameters, relative performance, orientation and cohesion. Our results suggest that the industry increased relative performance and cohesion over time, balancing company and customer perspectives, but regulatory compliance indicators lag those of other stakeholders.

2 - Branch-and-Price-and-Cut for Network Design with Reload Costs S. Raghavan, Associate Professor of Management Science, Robert H. Smith School of Business, University of Maryland, 4352 Van Munching Hall, College Park, MD, 20742, United States, [email protected], Luis Gouveia, Ioannis Gamvros
We present arc-flow and path-based models for network design problems with reload costs. We discuss the advantages and disadvantages of these approaches and present computational results on various problems.

4 - Evaluating Information Technology Investments on Bank Performance Yun Chieh Chin, Graduate Student, Department of Business Administration, Chang Gung University, Taoyuan, 333, Taiwan, [email protected]
Information technology has impacts on two main stages of bank operations, one is deposit service and the other is profit generation by investment using deposits. This research applies two-stage data envelopment analysis (DEA) to measure the operating efficiency of 32 commercial banks in Taiwan. The main findings show that older banks outperform newly established ones in each stage, and both at deposit and investing stages banks with large capitalization are more efficient than smaller ones.

3 - Regenerator Location Problem Si Chen, Univeristy of Maryland, Van Munching Phd #29, College Park, MD, 20742, United States, [email protected], S. Raghavan
We address the regenerator location problem in optical networks. In the network, every node is a terminal node and can be installed a regenerator. The objective is to determine a regenerator deployment so that all the nodes are connected and the number of the regenerators is minimized. A mixed integer program is presented. Three heuristics are proposed that solves large-sized networks in seconds. We also extend the regenerator location problem to generalized networks.

5 - Profit and Productivity of U.S. Class I Railroads Siew Hoon Lim, Assistant Professor of Economics, Department of Agribusiness & Applied Economics, North Dakota State University, 203B Morrill Hall, PO Box 5636, Fargo, ND, 58105, United States, [email protected], Knox Lovell
This paper examines how productivity changes and price changes have contributed to profit change in the railroad industry. We find that productivity improvements and an increased scale of production contributed to increases in profit, and that variation in operating efficiency had a mixed impact on profit. We also find that relative changes in rail rates and variable input prices exerted downward pressure on profit.

4 - An Exact Algorithm for Multistage Integer Multi-Commodity Flow Problems with Demand Uncertainty Ioannis Gamvros, Staff Consultant, ILOG, Inc., 1080 Linda Vista Avenue, Mountain View, CA, 94043, United States, [email protected], S. Raghavan
Multi-stage stochastic programming problems in which decision variables are integer in all stages are exceptionally hard and are typically approached with approximation algorithms. We propose a reformulation scheme and an associated branch-and-price-and-cut approach that can solve exactly stochastic multistage integer multi-commodity flow problems. We present computational results on a problem motivated from the telecommunications industry.

I SB52
Revenue Management Education and Material
Sponsor: Revenue Management & Pricing Sponsored Session
Chair: Ioana Popescu, Associate Professor of Decision Sciences, INSEAD, Boulevard de Constance, Fontainebleau, 77305, France, ioanaDOTpopescuATinseadDOTedu 1 - The Fjord Motor Custom Fleet Pricing Exercise Robert Phillips, Nomis Solutions, [email protected], J. Michael Harrison
Students are given data on 4000 past bids for fleet sales to police departments and corporations. Maximum likelihood is used to fit logit bid-response models at three levels of refinement. A second layer of Solver application gives the optimal pricing strategy under each bid-response model.

I SB51
Data Envelopment Analysis II
Contributed Session
Chair: Siew Hoon Lim, Assistant Professor of Economics, Department of Agribusiness & Applied Economics, North Dakota State University, 203B Morrill Hall, PO Box 5636, Fargo, ND, 58105, United States, [email protected] 1 - Interior Point Methods to Determine Output Prices in DEA Jose Dula, Associate Professor, Virginia Commonwealth University, School of Business, Richmond, VA, 23284, United States, [email protected], Marie-Laure Bougnol, Paul Rouse
Extreme efficient and inefficient DMUs with benchmarks on lower dimensional faces generate multiple optima in multiplier LPs. They are problematic as virtual weights to determine prices. IP methods identify analytical centers of the optimal face that satisfy SCSC. These solutions offer tangible advantages, e.g. uniqueness, well-defined rates of substitution, and proper envelopment. We report on results in the health care industry.

2 - MBA Revenue Management Course at Columbia University Costis Maglaras, Associate Professor, Columbia University, Graduate School of Business, 409 Uris Hall, 3022 Broadway, New York, NY, 10027, United States, [email protected], Garrett Van Ryzin
We present the structure of a full semester MBA elective course on Pricing and Revenue Management that we teach at Columbia, highlighting two new cases that we wrote recently.

2 - Linear Programming System Identification: Results for the General Nonnegative Parameters Case Marvin D. Troutt, Kent State University, Box 5190, Kent, OH, 44242-0001, United States, [email protected], Alan A. Brandyberry, Changsoo Sohn, Suresh K. Tadisina

3 - Teaching Revenue Management in the Carnegie Mellon Tepper School of Business MBA Program Nicola Secomandi, Assistant Professor, Carnegie Mellon University, Tepper School of Business, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected]
This talk illustrates the speaker’s experience teaching an MBA revenuemanagement elective at the Carnegie Mellon Tepper School of Business since 2004.

84

INFORMS PITTSBURGH— 2006
4 - Easy Profit - An Interactive Revenue Management Simulator Ioana Popescu, Associate Professor of Decision Sciences, INSEAD, Boulevard de Constance, Fontainebleau, 77305, France, ioanaDOTpopescuATinseadDOTedu
This interactive simulation enhances intuition about the profitability of inventory rationing decisions in RM. The user is in charge of optimizing revenues from the economy cabin of an aircraft. Business/leisure arrivals follow a stochastic process, which can be controlled by the educator. Static and dynamic booking limit strategies are tested, as well as sensitivity to input parameters. The material covers one or two sessions in a revenue/service/operations management course.

SB55

2 - Analyzing Capacity and Inventory Pooling Using Cooperative Game Theory Eda Kemahlioglu-Ziya, Kenan-Flagler Business School, University of North Carolina, CB # 3490, Chapel Hill, NC, 27599, United States, [email protected]
In this talk, we are going to talk about how cooperative game theory can be used to model and allocate savings resulting from pooling. We are going to give examples from inventory and capacity pooling. We will especially concentrate on the Shapley value as a solution approach. In the context of inventory pooling, we will talk about the behavior induced if Shapley value is used to allocate savings.

I SB53
Dynamic Models of Selling Mechanisms
Sponsor: Revenue Management & Pricing Sponsored Session
Chair: Jeremie Gallien, Associate Professor, Sloan School of Management, Massachusetts Institute of Technology, E53-389, 30 Wadsworth Street, Cambridge, MA, 02142, United States, [email protected] 1 - Adaptive Online Allocation Mechanisms for Single-Valued Domains David Parkes, Associate Professor, Harvard University, 33 Oxford Street, Cambridge, MA, 02138, United States, [email protected], Satinder Singh
We consider a model of adaptive online mechanisms for bidders with singlevalued types in a dynamic environment. Strong monotonicity and frugality are recognized as sufficient structural properties on allocation policies for dominantstrategy equilibrium, leading to adaptive, allocatively-efficient mechanisms.

3 - Inventory Ordering Consolidation: The Sensitivity of the Core Moshe Dror, Professor, University of Arizona, MIS, Eller College of Management, Tucson, AZ, 85721, United States, [email protected], Wei Cheng, Bruce C. Hartman
In a multi-item EOQ environment with two-part ordering cost composed of a constant shared cost and an item specific cost, inventory ordering consolidation cost has to be distributed among its items. It can be described as a cooperative game problem. In this talk we discuss the core of this game (a set of cost allocations that are fair) and its sensitivity to different problem parameters. The relation of concavity and inseparability of the game is examined computationally as well as other parameter properties.

I SB55
Supply Chain Logistics II
Contributed Session
Chair: Francisco Briones, Head of Department, Universidad Politecnica de Aguascalientes, Prol. Mahatma Gandhi Km. 2, Aguascalientes, 20280, Mexico, [email protected] 1 - A Multi-Echelon Inventory System with Shared Inventory Information Xiuhui Li, PhD, Nanyang Business School, Nanyang Technological University, S3-01B-73, Nanyang Avenue, Singapore, 639798, Singapore, [email protected], Qinan Wang
We consider a two-echelon supply chain consisting of a supplier and multiple identical buyers. Demand at the buyers follows Poisson process. We develop the optimal inventory replenishment policy for the supplier when on-line information on the buyers’ inventory status is available. This analysis provides an exact evaluation of the value of on-line information for the system.

2 - Online Auctions for Market Research on Consumers’ Willingness to Pay Matthew Potoff, PhD Candidate, Ross School of Business, University of Michigan, 701 Tappan Street, Ann Arbor, MI, 48109, United States, [email protected], Damian Beil
We consider a manufacturer selling a short life-cycle product over a finite selling horizon. The manufacturer is uncertain about consumers’ willingness to pay. Initially items are sold via auction and the manufacturer’s priors about willingness to pay are updated. Using a DP approach we explore when the manufacturer should select a price and switch to a mass-market, fixed price channel. Consumer product adoption follows a diffusion process over the horizon.

3 - Optimal Dynamic Auctions Mallesh Pai, Doctoral Student, Managerial Economics and Decision Sciences, Kellogg School of Management, 2001 Sheridan Road, Northwestern University, Evanston, IL, 60208, United States, [email protected], Rakesh Vohra
We consider a dynamic auction motivated by the single-leg, multi-period revenue management problem. A monopolist sells C indivisible units over T time periods. A buyer’s arrival time, desired quantity, valuation and time by which the purchase must be made are private information. We derive the revenue maximizing Bayesian incentive compatible mechanism.

2 - Information Interpretation in the Supply Chain: Role of Human Decision Making on the Bullwhip Effect David Cantor, Visiting Assistant Professor, Smeal College of Business, Penn State University, Supply Chain and Information Systems, 460 Business Building, University Park, PA, 16802, United States, [email protected], John Macdonald
We investigate the role of cognitive thinking styles in the supply chain. We theoretically draw on construal theory to test our model that the type of information given as well as cognitive thinking style contributes to the bullwhip effect. A key finding is that abstract thinkers filter local information better than concrete thinkers. However, both types of thinkers are overwhelmed with complete information, and subsequently make poor decisions which leads to the bullwhip effect.

I SB54
Inventory Games in Supply Chain
Cluster: Operations and Marketing for Emerging Markets Invited Session
Chair: Moshe Dror, Professor, University of Arizona, MIS, Eller College of Management, Tucson, AZ, 85721, United States, [email protected] 1 - Stable Cooperation in Periodic Review Inventory Situations Ana Meca, Professor, Universidad Miguel Hernandez, Operations Research Center, Elche, Spain, [email protected], Luis A. Guardiola, Justo Puerto
This paper examines coordination in periodic review inventory models in which agents can share ordering, inventory holding, and backlogging resources. We analyze the resulting cooperative game and show that the cooperation is stable (no group of agents would like to leave the system). A family of fair divisions of costs is introduced and the relation to other inventory coordination models from the literature is studied.

3 - Optimization Opportunities to Improve Supply Chain Efficiency: A Practitioner’s Experiences Rajiv Saxena, Director, Supply Chain Engineering, APL Logistics, 1111 Broadway, Oakland, CA, 94607, United States, [email protected]
Different companies with global supply chains are looking towards their logistics providers to help them in successfully dealing with their supply chain challenges. We discuss case studies on how we created significant efficiencies in different aspects of global supply chains of our customers through optimally engineered solutions.

4 - Inventory Management for Customers with Alternative Lead-Time Choices Haifeng Wang, The Chinese University of Hong Kong, Room 615, Mong M. W. Engineering Building, Shatin, N.T., Hong Kong, [email protected], Houmin Yan
This paper considers an inventory model with alternative lead-time choices. We obtain the optimal dynamic inventory commitment policy about when to deliver; and prove the optimal inventory replenishment policy to be a base-stock type. Further, we obtain that the inventory commitment policy outperforms an inventory rationing policy, and find the robustness of our policy. We use the customer choice model to characterize the risk pooling, demand induction and cannibalization effects.

85

SC01

INFORMS PITTSBURGH — 2006
2 - Parallel Asynchronous Derivative-Free Optimization with Nonlinear Constraints Joshua Griffin, Limited Term Employee, Sandia National Laboratories, Mail Stop 9159, PO Box 969, Livermore, CA, 94551, United States, [email protected], Tamara Kolda
We present a globally convergent, derivative-free method for nonlinear programming based on generating set search (GSS). We explore several options for handling nonlinear constraints, including an augmented Lagrangian approach. Linear constraints are straightforward, requiring only conforming search directions. All methods are implemented in parallel and asynchronously using APPSPACK. We present extensive test results using the CUTEr test set as well as real-life applications.

5 - Petri Nets for Description of Batch Processes in Function of Coordinator Control Recipe Francisco Briones, Head of Department, Universidad Politecnica de Aguascalientes, Prol. Mahatma Gandhi Km. 2, Aguascalientes, 20280, Mexico, [email protected]
Petri Nets can be adapted to describe the discrete and dynamical behavior of a Batch Process and how it can be used to formulate control strategies supported by a Coordinator Control Recipe in terms of ISA S88.02 norm.

Sunday, 1:30pm - 3:00pm
I SC01
Algorithms for Discrete Stochastic Optimization Models
Sponsor: Optimization/ Network and Combinatorial Optimization Sponsored Session
Chair: David Shmoys, Professor, Cornell University, School of OR&IE, Ithaca, NY, 14853, United States, [email protected] 1 - Data-Driven Convex Optimization with Applications to Supply Chain Management Paat Rusmevichientong, Assistant Professor, Cornell University, 221 Rhodes Hall, Ithaca, NY, 14850, United States, [email protected], W. Tim Huh
We study non-parametric stochastic inventory planning systems with lost sales and censored demand under stationary and non-stationary settings. We propose adaptive inventory policies that generate a sequence of ordering decisions over time. The decision in each period depends only on historical sales data of the past. Our adaptive algorithms converge to the optimal, and the average expected cost during the first T periods differs from the optimal cost by at most O( 1 / \sqrt{T} ).

3 - JEGA: A Tool for Multi-Objective Optimization John Eddy, Senior Member of the Technical Staff, Sandia National Laboratories, Mail Stop 1125, PO Box 5800, Albuquerque, NM, 87185-1125, United States, [email protected]
JEGA is a software package that implements a multi-objective genetic algorithm (MOGA) for solution to multi-objective problems. JEGA is highly configurable and provides a rich set of algorithmic components for tailored optimization. It has been used successfully on many problems by users both internal and external to Sandia.

4 - OPT++: An Object-Oriented Toolkit for Nonlinear Optimization Patricia Hough, Principal Member of Technical Staff, Sandia National Laboratories, PO Box 969, MS 9159, Livermore, CA, 94551, United States, [email protected], Pamela Williams, Ricardo Oliva, Juan Meza
We describe OPT++, a C++ library for nonlinear optimization. The design is predicated on distinguishing between an algorithm-independent class hierarchy for nonlinear optimization problems and a class hierarchy for nonlinear optimization methods. The interface is designed for ease of use and extensibility to new algorithms. Several nonlinear optimization algorithms have been implemented, and example problems demonstrate the advantages of a common interface in comparing multiple algorithms.

2 - Transforming Stochastic Dynamic Programs to Yield a FPTAS, with an Application Frans Schalekamp, Cornell University, School of OR&IE, Ithaca, NY, 14853, United States, [email protected], David Shmoys
Methods to transform (certain) Stochastic Dynamic Programs to yield a fully polynomial-time approximation scheme (FPTAS) are proposed. The methods are generalizations of a method proposed by Halman et al. Finally an application in option pricing is presented.

I SC03
Approximate Dynamic Programming
Sponsor: Optimization/ Linear Programming and Complementarity Sponsored Session
Chair: Daniela Pucci de Farias, MIT, [email protected] 1 - Using Stochastic Approximation to Compute Optimal Base-Stock Levels in Inventory Control Problems Sumit Kunnumkal, Graduate Student, Cornell University, School of ORIE, 295 Rhodes Hall, Ithaca, NY, 14850, United States, [email protected], Huseyin Topaloglu
We consider inventory control problems where base-stock policies are known to be optimal and propose stochastic approximation methods to compute the optimal base-stock levels. Existing methods guarantee convergence, but not necessarily to the optimal base-stock levels. In contrast, we prove that the iterates of our methods converge to the optimal base-stock levels. Computational experiments indicate that our methods provide significantly better solutions than existing methods.

3 - Provably Near-Optimal Sampling-Based Policies for Stochastic Inventory Control Models Retsef Levi, Assistant Professor, Massachusetts Institute of Technology, Sloan School of Management, BDG E53-389, 30 Wadsworth Street, Cambridge, MA, 02139, United States, [email protected], Robin Roundy, David Shmoys
Consider the classical newsvendor model and its multiperiod extension, but under assumption that the exact demand distributions are not known. The only thing available is a set of samples drawn from the true distributions. We present sampling-based policies computed only based on samples and bound the number of samples required to guarantee that their expected cost is arbitrarily close to the optimal cost computed based on the true distributions. The bounds apply to any demand distribution.

I SC02
Joint Session Open-Source/ICS: Open-Source Optimization Packages for Engineering Design
Cluster: Open-Source Software: Open Source, Open Standards, Open Data, INFORMS Computing Society Invited Session
Chair: William Hart, Distinguished Member of Technical Staff, Sandia National Laboratories, Mail Stop 1110, PO Box 5800, Albuquerque, NM, 87185-1110, United States, [email protected] 1 - DAKOTA: A Toolkit for Engineering Design William Hart, Distinguished Member of Technical Staff, Sandia National Laboratories, Mail Stop 1110, PO Box 5800, Albuquerque, NM, 87185-1110, United States, [email protected], Michael Eldred, Anthony Giunta
We summarize capabilities and applications of the DAKOTA toolkit for high performance computers. It provides a flexible interface between simulation software and algorithms for optimization, uncertainty quantification, parameter estimation, design of experiments, and sensitivity analysis. DAKOTA enables research in multilevel parallel computing, mixed integer nonlinear programming, surrogate-based optimization, optimization under uncertainty, and simultaneous analysis and design.

2 - Serial Multi-Echelon Systems with Economies of Scale via Approximate Dynamic Programming Diego Klabjan, Associate Professor, University of Illinois at Urbana-Champaign, 1206 West Green Street, Urbana, IL, 61801, United States, [email protected]
We consider the serial multi-echelon system with transportation economies of scale. Since optimal policies are not known, we apply approximate dynamic programming to compute policies. We experiment with several value function approximations.

3 - Decentralized Approximate Dynamic Programming and Resource Allocation for Dynamic Networks of Agents Daniela Pucci de Farias, [email protected], Hariharan Lakshmanan
We consider stochastic control systems involving teams of agents communicating through a network with dynamic topology. We propose a decentralized approximate dynamic programming scheme and show that it can be viewed as a resource allocation problem. Motivated by this observation, we develop a fully decentralized and asynchronous algorithm for a general class of resource allocation problems that is provably convergent under mild assumptions on the structure of communication among agents.

86

INFORMS PITTSBURGH— 2006
4 - An Optimal Approximate Dynamic Programming Algorithm to a Mutual Fund Problem Juliana Nascimento, PhD Candidate, Princeton University, Olden Street, Equad Room E-307, Princeton, NJ, 08544, United States, [email protected], Warren B. Powell
We present an approximate dynamic programming algorithm to optimize the level of cash that is held by a mutual fund in order to meet redemption requests from investors. Demand for redemptions and shortfall/surplus costs are stochastic. The algorithm uses a pure exploitation scheme and does not require knowledge of the distribution of any of the random variables. A proof of convergence to an optimal policy is obtained exploring structural properties of the problem.

SC06

I SC05
Stochastic Programming
Sponsor: Optimization/ Stochastic Programming Sponsored Session
Chair: Shabbir Ahmed, Associate Professor, Georgia Institute of Technology, 765 Ferst Drive, Atlanta, GA, 30068, United States, [email protected] 1 - Fire-Fighting Resource Allocation for Wildfire Containment Using Stochastic Integer Programming Lewis Ntaimo, Professor, Department of Industrial & Systems Engineering, Texas A&M University, 3131 TAMU, College Station, TX, 77840-3131, United States, [email protected], Won Ju Lee
Recent catastrophic wildfires have highlighted the need for new models to assist fire managers in containment decision-making under uncertainty. This talk presents a stochastic integer programming approach to the difficult problem of initial attack for fire containment. The problem involves determining the optimal mix of fire-fighting resources to contain a fire under uncertainty in fire growth. Deterministic models assume that fire growth is known and are therefore, inadequate for this problem.

I SC04
Laboratory Experiments in Supply Chain and Operations Management
Sponsor: Manufacturing & Service Operations Management Sponsored Session
Chair: Elena Katok, Associate Professor, Penn State University, 465 Business Building, University Park, PA, 16802, United States, [email protected] 1 - On the Perception of Supply Contract Mirko Kremer, University of Mannheim, Schloss, S 231, Mannheim, BW, Germany, [email protected]
Contradicting the normative benchmark, recent experimental research indicates that actual order behavior is differs under a revenue sharing and a buy back contract which are deemed to be technically equivalent in a simple supply chain setting with a supplier selling to a newsvendor. We extend on this matter by considering choice between contracts rather than the behavior under them. Our results are largely an issue of decision framing but can be neatly tied to established behavioral theory.

2 - A Multilevel Resource Allocation Problem for Stream Processing Systems Li Zhang, Manager, Systems Analysis and Optimization, IBM TJ Watson Research Center, 19 Skyline Drive, PO Box 704, Hawthorne, NY, 10532, United States, [email protected], Hanhua Feng
Data streaming applications are becoming popular due to the rapid development in sensor networks, multimedia streaming, and online data mining, etc. The incoming data streams often fluctuate at a finer time scale compared to the resource allocation decisions. We employ a multilevel resource allocation scheme to account for the different time scale and the randomness in the stream arrival rates. Stochastic programming approaches are presented to obtain the optimal resource allocation solutions.

2 - Impact of Fairness on Supply Chain Contracts: A Laboratory Study Diana Wu, University of Kansas, School of Business, Lawrence, KS, United States, [email protected]
The topic of supply chain coordination has been much studied. While theoretical properties of various contracting mechanisms have been studied extensively, less is known about their practical performance. Using a controlled setting of an experimental laboratory, we investigate the impact of supply chain partners’ equity concerns over the profit and risk divisions on the performance of different contracting mechanisms.

3 - Solving Multistage Stochastic Linear Programs on the Computational Grid Jierui Shen, PhD Candidate, Lehigh University, Industrial and Systems Engineering, 200 W. Packer Avenue, Bethlehem, PA, 18015, United States, [email protected], Jeff Linderoth
We describe a solver capable of tackling large-scale multistage stochastic linear programs (MSLP). The solver is based on a nested-decomposition (ND) algorithm, and we study how to combine the algorithm with the power of grid computing. Computational results addressing the challenges of effectively implementing ND on a grid are given. Using our solver, we also study the value obtained in solving MSLP as opposed to alternative methods for dealing with uncertainty.

3 - Inventory Service Level Agreements as Coordination Mechanisms Doug Thomas, Penn State University, 463 Business Building, University Park, PA, United States, [email protected], Andrew Davis, Elena Katok
A supplier balancing inventory cost and profit margin may choose a stock level that maximizes her profitability but is sub-optimal from the view of the retailer and the overall supply chain. To address this double marginalization problem, a retailer may introduce a service level agreement, offering a bonus for meeting a predetermined fill rate over a specified (finite) horizon. We experimentally investigate how such service level agreements affect supplier stocking decisions.

4 - Strong MIP Formulations of Chance Constrained Linear Programs with Random Right-hand Side James Luedtke, PhD Student, Georgia Institute of Technology, 765 Ferst Drive, Atlanta, GA, 30332, United States, [email protected], Shabbir Ahmed, George Nemhauser
We present strong mixed-integer programming formulations for the chance constrained linear program with finite sample space and random RHS. We develop valid inequalities and an extended formulation based on a single row relaxation. Computational results indicate this approach can be used to solve large scale problems, with the size of the sample space in the thousands and dimension of the random RHS in the hundreds.

4 - Supply Chain Coordination with Risk-Averse Players: A Behavioral Analysis Julie Niederhoff, Washington University in St Louis, Olin School of Business, One Brookings Drive; Campus Box 1133, St Louis, MO, 63130, United States, [email protected], Panos Kouvelis
We study supply chain coordination under risk-aversion. Players are measured for risk preferences and then act as profit-maximizing suppliers or retailers to determine which mechanisms are most effective. We use the two-member newsvendor supply chain. We use traditional theory of risk-neutral and riskaverse players as a benchmark to measure actual performance.

I SC06
Constraint Programming Tools
Sponsor: INFORMS Computing Society/Constraint Programming Sponsored Session
Chair: Laurent Michel, Assistant Professor, Computer Science & Engineering Department, University of Connecticut, Storrs, CT, 06269, United States, [email protected] 1 - Gecode - A Generic Constraint Development Environment Christian Schulte, Associate Professor, KTH - Royal Institute of Technology, KTH/ICT/ECS, Isafjordsgatan 39, Electrum 229, Kista, 16440, Sweden, [email protected]
Gecode (www.gecode.org) is an open, free, portable, accessible, and efficient environment for developing constraint-based systems and applications. In this talk I will focus on what makes Gecode radically open for programming (variables, constraints, search) and why this matters for developers and users.

87

SC07

INFORMS PITTSBURGH — 2006
3 - Operations and Supply Management - The Core Robert Jacobs, Professor of Operations Management, Indiana University, 1309 East Tenth Street, Bloomington, IN, 47405, United States, [email protected]
The new book concisely covers “core” operations management topics in 13 chapters. A theme that is carried throughout is that success for companies today requires managing the entire supply flow, from the sources of the firm, through the value added processes of the firm, and on to the customers of the firm. The first edition book will be available from McGraw-Hill during November, 2006.

2 - Recent Developments in Constraint Programming at ILOG Paul Shaw, [email protected], Jean-Francois Puget, Jean-Charles Regin
ILOG Solver is the leading commercial constraint programming tool, used by industry and academia alike. In this talk, we present the recent developments in constraint programming at ILOG with respect to ease of use, search procedures and solver performance.

3 - The CHOCO Constraint Solver Hadrien Cambazard, Ecole des Mines de Nantes, La Chantrerie - 4, rue Alfred Kastler, BP 20722, Nantes Cedex 3, France, [email protected], Narendra Jussien, François Laburthe, Guillaume Rochart
The CHOCO constraint solver is an emanation of a French academic group. Choco is a java library for constraint satisfaction problems, constraint programming, and explanation-based constraint solving. It is built on an eventbased propagation mechanism with backtrackable structures. Choco also provides explanations as both an analysing and solving tool for constraint programming, uncommon search mechanisms including the decision-repair and the logical Benders decomposition schemes, etc.

4 - Randomizing and Automating Homework and Exam Grading Jay Heizer, Jones Prof of Bus Admin, Texas Lutheran University, Dept. of Business, Seguin, Tx, 78155, United States, [email protected]
Software that randomizes and automates both problems and values for homework and exam delivery in now available for the Operations Course. The system also grades the material and posts to a grade book so faculty can improve their teaching, have more time for students, and research, and life.

4 - Differentiability in Constraint-Based Local Search Laurent Michel, Assistant Professor, Computer Science & Engineering Department, University of Connecticut, Storrs, CT, 06269, United States, [email protected], Pascal Van Hentenryck
Differentiable invariants are generic structures that maintain the value of arbitrary expressions and their variable gradients. They support differentiation to evaluate the effect of moves performed within a local search procedure. The benefits of differentiable invariants are illustrated on a number of applications which feature complex, possibly reified, expressions and whose models are essentially similar to their CP counterparts. Experimental results demonstrate their practicability.

I SC08
Sponsor: Technology Management Sponsored Session
Chair: Nile Hatch, Assistant Professor, Marriott School, Brigham Young University, 790 TNRB, Provo, UT, 84602, United States, [email protected] 1 - Exploration and Exploitation in Complex Networks: Learning Rates and Interpersonal Networks Melissa Schilling, Associate Professor of Management, Stern School of Business, New York University, 40 West Fourth Street, New York, NY, 10012, United States, [email protected]
The rate at which superior solutions spread through an interpersonal network is strongly influenced by the learning rate utilized by individuals and the underlying network structure. We find that very low levels of random linking increase performance, but higher levels of random linking lower performance. Small-world network properties can enhance organizational learning by enabling short path lengths and semi-isolated pockets of heterogeneous knowledge to exist simultaneously in a network.

5 - Differentiable Invariants for Constraint-Based Local Search Pascal Van Hentenryck, Professor of Computer Science, Brown University, Box 1910, 115 Waterman Street, 4th Floor, Providence, RI, 02912, United States, [email protected], Laurent Michel
This talk proposes the concept of differentiable invariants as a natural abstraction to express constraint-based local search algorithms. Differentiable invariants maintain the value of complex expressions incrementally, derive and maintain gradients for each variable, and allow efficient evaluations of local moves. Their expressive power and efficiency are demonstrated on a number of resource allocation applications.

2 - Cross-Training, Social Identity and Knowledge Sharing Enno Siemsen, Assistant Professor of Business Administration, University of Illinois at Urbana-Champaign, Department of Business Administration, 350 Wohlers Hall, 1206 S. Sixth Street, Champaign, IL, 61820, United States, [email protected], Sridhar Balasubramanian, Aleda Roth
This research explores the relationship between cross-training and interemployee knowledge sharing. Cross-training can increase knowledge sharing, for example by enabling better communication between employees. On the other hand, cross-training increases the functional similarity between employees, which in turn can increase the competition between them, thereby reducing knowledge sharing behavior. We explore these relationships empirically using survey data from three different companies.

I SC07
Supply Chain Management Books
Cluster: New Books Invited Session
Chair: David Woodruff, Professor, University of California Davis, Graduate School of Management, 155 AOB 4, Davis, CA, 95616, United States, [email protected] 1 - Introduction to Computational Optimization Models for Production Planning in a Supply Chain David Woodruff, Professor, University of California - Davis, Graduate School of Management, 155 AOB 4, Davis, CA, 95616, United States, [email protected], Stefan Voss
Managers and information technology professionals need to have an understanding of computational optimization models for production planning in a supply chain. This book provides an accessible introduction to the subject. We develop the terminology and concepts needed to understand the important issues.

3 - Dissecting Organizational Knowledge: Multimodal Apprenticeship in Vascular Surgery Curtis LeBaron, Associate Professor, Brigham Young University, 790 Tanner Building, Marriott School, Provo, UT, 84602, United States, [email protected]
This is a video-based study of instruction within a surgical team. The attending surgeon, who is most expert, provides subtle visible and vocal prompts for the work of the resident, who is a novice. Such “scaffolding” behaviors are a form of practice that emerge organically within surgical activity, supporting the novice’s performance as needed and to the extent needed, disappearing into the folds of interaction as the resident develops expertise.

2 - The New Krajewski-Ritzman-Malhotra Operations Management Text from Prentice-Hall Lee Krajewski, Professor, University of Notre Dame, 363B Mendoza College of Business, Notre Dame, IN, 46556, United States, [email protected]
The new eighth edition of Operations Management: Processes and Value Chains is slimmer and focuses more on the use of operations management for all business majors. The text features a new video interactive-case series about Starwood Hotels and reinforces the use of operations management for non manufacturing processes throughout the text. Major changes have been made to project management, supply chain strategy, location, sales and operations planning, and constraint management.

4 - Fractal Knowledge and Technological Change in Semiconductors and Disk Drives Roger Bohn, Professor, University of California - San Diego, 9500 Gilman Drive #0519, IR/PS, La Jolla, CA, 92093-0519, United States, [email protected]
Knowledge can be modeled as a causal graph which grows as more is learned. Such graphs are fractal: parent-child relationships become complex subgraphs as more is learned. These patterns are illustrated by semiconductor lithography and hard disk drive head design. As product generations get smaller (Moore’s Law), new physical relationships become large enough to be important. Also, incremental and radical technical change can be measured by how the graph changes.

88

INFORMS PITTSBURGH— 2006 I SC09
Management of Innovation & New Product Development in Complex Systems Environments
Cluster: New Product Development Invited Session
Chair: Edward Anderson, Professor, University of Texas, McCombs Business School, 1 University Station B6500, Austin, TX, 78733, United States, [email protected] 1 - The Butterfly Effect: Systemic and Behavioral Risks in Distributed Innovation Nitin Joglekar, Associate Professor, Boston University, 525 Commonwealth Avenue, Boston, MA, 02215, United States, [email protected], Edward Anderson
We explore distributed innovation decisions from systems thinking perspective. This approach enables us to link the architectural, behavioral, competitive, and demand risks though feedback mechanisms. We argue that decisions underlying such innovation structures are susceptible to the butterfly effect: path dependent amplification of small uncertainties due to feedback and dynamic complexity can lead unintended consequences such as turf building and tipping the balance within a firm’s portfolio.

SC11

I SC10
Software Demonstration
Cluster: Software Demonstration Invited Session
1 - StatPoint, Inc. - Synchronized Data Analysis Using STATGRAPHICS Centurion and STATGRAPHICS Mobile Neil Polhemus, StatPoint, Inc., 2325 Dulles Corner Blvd., Ste. 500, Herndon, VA, 20171, [email protected], Seth Wyatt
StatPoint will demonstrate the interaction between STATGRAPHICS Centurion, its award-winning desktop statistical analysis package, and STATGRAPHICS Mobile, the first serious statistical software package for Pockets PCs and other handheld devices. The presentation will include a discussion of data synchronization and an illustration of distributed forecasting using ARIMA models.

2 - Salford Systems - RandomForests and TreeNet/MART Mikhail Golovnya, Salford Systems, 4740 Murphy Canyon Rd., #200, San Diego, CA, 92108, [email protected]
Salford Systems will demo both RandomForests and TreeNet/MART, Leo Breiman’s and Jerome Friedman’s newest advances to data mining. Both RandomForests and TreeNet enable the modeler to construct predictive models of extraordinary accuracy. Models are built up gradually through a large collection of small trees, each of which improves on its predecessors through an errorcorrecting strategy.

2 - Product Development Competitions as an Integrated Product Development Strategy Saurabh Bansal, Graduate Student Teaching Assistant, McCombs School of Business, University of Texas - Austin, 1 University Station B6500, Austin, TX, 78712, United States, [email protected], Edward Anderson
We present a model that considers a two-pronged strategy for new product development (NPD) at a firm: 1) Product development through an in-house team, and 2) Product development through competitions. The solutions received from the competition expedite the NPD learning process while the in-house team develops technical details to implement these solutions. Results indicate that competitions can expedite the NPD process when the technical knowledge about the new project is sparse in the firm.

I SC11
OR in SpORts II
Sponsor: OR in Sports Sponsored Session
Chair: Joel Sokol, Assistant Professor, Georgia Institute of Technology, 765 Ferst Drive NW, Industrial & Systems Engineering, Atlanta, GA, 30332-0205, United States, [email protected] 1 - “A Different Game”? An Analysis of Scoring in the Baseball World Series and a Markovian Model Vijay Mehrotra, Assistant Professor, San Francisco State University, College of Business, 1600 Holloway Avenue, San Francisco, CA, 94610, United States, [email protected]
Many observers assert that a team should focus on “manufacturing runs” by using tactics such as sacrifice bunting and base stealing. We use statistical methods along with data from the past twenty World Series to examine two key questions. First, we investigate whether World Series games feature (statistically) significantly lower scoring. Secondly, we utilize a Markovian model to examine the value of sacrifice bunts and stolen bases and the implications for managerial tactics.

3 - Adaptive Infrastructure: Learning from Open Source Communities and Toyota Paul Carlile, Associate Professor, Boston University, 595 Commonwealth Avenue, Boston, MA, 02215, United States, [email protected], Karim Lakhani
We compare two very innovative, but different distributed innovation systems: Open Source Software communities and Toyota. The adaptive value of their infrastructure consists of actors and artifacts and their distributed capability to identify the consequences of novel information and then make collective changes accordingly.

4 - Managing the Life Cycle Mismatch Problem James Bradley, Associate Professor, The College of William & Mary, 118C Tyler Hall, Mason School of Business, Williamsburg, VA, 23187-8795, United States, [email protected], Hector Guerrero
“Life-cycle mismatch” occurs when component life cycles are shorter than the life cycles of products in which they are used. Potential responses to premature component obsolescence are expensive (e.g., product redesign, life-time component buys) and should be taken into account in the initial design of a product. We evaluate when a durable design, which is less prone to component obsolescence than a nondurable design, is optimal.

2 - Analyzing College Football Overtime Strategy Rick Wilson, Professor, Department of Management Science and Information Systems, Oklahoma State University, 408 BUS, Spears School of Business, Stillwater, OK, 74078, United States, [email protected], Peter Rosen
Division I-A college football adopted “alternating possession” overtime rules in 1996. Over this 10 year period, very few teams have varied from conventional wisdom that indicates a teams’ chances of winning increase if you are on defense first. We analyze game data over the past decade and show that while there is some evidence of an advantage for this strategy, it is not as pronounced as many think.

5 - Managing Risk in Alternative Energy Product Development Geoffrey Parker, Associate Professor, Tulane University, Freeman School of Business, New Orleans, LA, 70118, United States, [email protected], Edward Anderson
Developing alternative energy products is inherently risky because of the volatile nature of oil prices. This volatility stems from: (1) Cycles in oil exploration, and (2) Information asymmetry because energy firms with access to the most fruitful sites for discovery of new reserves may be underestimating them due to technological or geopolitical factors. This leads us to discuss the impact of these price risks on alternative energy product development and how these risks might be mitigated.

89

SC12 I SC12

INFORMS PITTSBURGH — 2006
Science, School of Business and Management, University of San Francisco, 2130 Fulton Street, San Francisco, CA, 94117, United States, [email protected]
INFORMS Case & Teaching Materials peer reviews and publishes teaching cases. We will discuss our existing strategic cases, our new crop of cases, our new editorial board and reviewing system, and opportunities to submit, review, and serve as associate editor.

Nicholson Student Paper Prize Competition, I
Cluster: Nicholson Student Paper Prize Invited Session
Chair: Julie Swann, Assistant Professor, Georgia Insititue of Technology, Industrial and Systems Engineering, Atlanta GA 30332, United States, [email protected] 1 - Generalized Mixed Integer Rounding Valid Inequalities Kiavash Kianfar, PhD Candidate, Department of Industrial Engineering, North Carolina State University, 458 Daniels Hall, Campus Box 7906, North Carolina State University, Raleigh, NC, 27695, United States, [email protected], Yahya Fathi
We present new families of valid inequalities for (mixed) integer programming (MIP) problems. These valid inequalities are based on a generalization of the 2step mixed integer rounding (MIR), proposed by Dash and Günlük (2006). We prove that for any positive integer n, n facets of a certain (n+1)-dimensional mixed integer set can be obtained through a process which includes n consecutive applications of MIR. We then show that for any n, the last of these facets, the n-step MIR facet, can be used to generate a family of valid inequalities for the feasible set of a general (mixed) IP constraint, the n-step MIR inequalities. The Gomory Mixed Integer Cut and the 2-step MIR inequality of Dash and Günlük (2006) are simply the first two families corresponding to n=1,2, respectively. The n-step MIR inequalities are easily produced using closed-form periodic functions, which we refer to as the n-step MIR functions. None of these functions dominates the other on its whole period. Moreover, for any n, the nstep MIR inequalities define new families of facets for the finite and the infinite group problems.

4 - Integrating Ethics into the OR/MS Classroom Matthew Drake, Assistant Professor, Duquesne University, A.J. Palumbo School of Business, Donahue Graduate School of Business, Pittsburgh, PA, 15282, United States, [email protected], David Wasieleski, James Weber
The importance of ethics instruction has been recognized by accreditation boards and many companies in recent years. Many OR/MS instructors, though, have yet to integrate ethics into their classrooms, partly because they do not think they have adequate training in the subject. We discuss ways in which even the most ethically-novice instructors can introduce ethics topics into their classrooms.

I SC14
Quality Monitoring
Contributed Session
Chair: Min Shi, Assistant Professor, Marist College, School of Management, 3399 North Road, Poughkeepsie, NY, 12601, United States, [email protected] 1 - Online Monitoring Network Characteristics Using Flexible Probing Experiments Xiaodong Yang, Graduate Student, The University of Michigan, 1205 Beal Avenue, Ann Arbor, MI, 48109, United States, [email protected]
Large-scale network monitoring has gained increasing attention, especially in light of stringent QoS requirements of real-time applications. In this talk, after discussing some change models, efficient methods for estimating QoS parameters such as packet loss rates and delay will be examined. Various control chart procedures will be compared in terms of the power of detection. This is joint work with Drs Vijay Nair and George Michailidis.

2 - The G/GI/N Queue in the Halfin-Whitt Regime Josh Reed, Georgia Institute of Technology, School of Industrial and Systems Engineering, 765 Ferst Drive NW, Atlanta, GA, 30332, [email protected]
The many-server, heavy traffic limit for the GI/M/N queue was first discovered by Halfin and Whitt in 1981. Since then, Halfin and Whitt’s results have been used extensively in research on large scale telephone call centers. However, the extension of Halfin and Whitt’s results to general service time distributions has remained an open problem. In this talk, we will consider the G/GI/N queue in the Halfin-Whitt regime. Our first result will be to obtain a one-dimensional heavy traffic limit for the properly centered and scaled queue length process. We will then proceed to establish an equivalent characterization of our limiting process involving the renewal function associated with the service time distribution. Finally, we will conclude by discussing several interesting directions for future research.

2 - Analysis of Signal-Response Systems Using GEE and GLMM Shilpa Gupta, Graduate Student, Arizona State University, Mail Code 5906, Department of Industrial Engineering, Tempe, AZ, 85287, United States, [email protected], Connie Borror, Douglas Montgomery
We illustrate the application of generalized estimating equations and the generalized linear mixed model approach to robust design and analysis of signal response systems. We propose a split-plot approach to the signal response system, characterized by two variance components - within-profile variance and between-profile variance. We demonstrate that decomposing the system variation in this manner leads to more precise point estimates with shorter confidence intervals.

3 - On Bid Price Controls for Network Revenue Management Baris Ata, Assistant Professor, Kellogg School of Management, Northwestern University, 2001 Sheridan Road, Evanston, IL, 60208, United States, [email protected], Mustafa Akan
We consider a network model of revenue management and propose a novel class of bid-price control mechanisms. Under mild assumptions, we show that the proposed bid-price control mechanism is optimal, and that the optimal bid prices form a martingale. Optimality and martingale property of bid-prices are robust to information distortions. In particular, we show that under semi-continuous information structures, there exist near-optimal predictable bid-price policies having the martingale property.

3 - A Change Point Model for the Exponential Family Marcus Perry, Assistant Professor of Operations Research, Air Force Institute of Technology, 2950 Hobson Way, Building 641, Wright-Patterson AFB, OH, 45433-7765, United States, [email protected]
Many change point models used in SPC are designed under the assumption that a single disturbance is present. However, this assumption is often unreasonable as many processes are susceptible to more complicated disturbances. In this research a change point model is proposed for linear disturbances in the natural location parameter of exponential densities.

I SC13
Teaching OR/MS with Cases
Sponsor: Education (INFORM-ED) Sponsored Session
Chair: Matthew Drake, Assistant Professor, Duquesne University, A.J. Palumbo School of Business, Donahue Graduate School of Business, Pittsburgh, PA, 15282, United States, [email protected] 1 - Case Learning Cliff Ragsdale, Bank of America Professor, Virginia Polytechnic Institute and State University, 1007 Pamplin Hall, Blacksburg, VA, 24061, United States, [email protected]
This presentation focuses on lessons learned from and about teaching with cases.

4 - Economic and Economic-Statistical Design of a Multivariate Bayesian Control Chart Zhijian Yin, PhD Student, University of Toronto, 5 King’s College Road, Toronto, ON, M5S 3G8, Canada, [email protected], Viliam Makis
The paper presents a cost model for a MVBCCH for a long production run. The objective is to determine the control chart parameters, namely the control limit, sample size, and the sampling interval minimizing the total expected average cost. The average cost function is derived using renewal theory. Additional statistical constraints guaranteeing the average run length performance are also considered.

2 - Teaching Management Science with Cases Greg Zaric, Professor, Ivey School of Business, University of Western Ontario, London, ON, Canada, [email protected]
This talk covers the use of cases for teaching management science. The talk will briefly touch upon the following issues: classroom management and encouraging class participation; evaluating student performance; choosing cases for classes and exams; and writing good cases.

3 - INFORMS Case & Teaching Materials Thomas Grossman, Associate Professor of Information & Decision

90

INFORMS PITTSBURGH— 2006
5 - Person-Fit Measurement Based on Adaptive Statistical Process Control Min Shi, Assistant Professor, Marist College, School of Management, 3399 North Road, Poughkeepsie, NY, 12601, United States, [email protected], Ronald Armstrong, Zachary Stoumbos
Person-fit measurement refers to statistical methods used to evaluate the fit of response patterns to tem response theory model. In this paper, a new class of cumulative sum (CUSUM) schemes based on log-likelihood-ratio statistics is developed to detect aberrant response patterns in linear tests where the dichotomous responses are independent. Extensive simulations are performed to compare the proposed CUSUM schemes with selected person-fit statistics in the literature.

SC17

I SC16
Revenue Management and Pricing under Various Market Models
Cluster: Dynamic Pricing and Forecasting Invited Session
Chair: Nicola Secomandi, Assistant Professor, Carnegie Mellon University, Tepper School of Business, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected] 1 - Comparing Price and Quantity Competition in Oligopolies with Differentiated Products Amr Farahat, Johnson School, Cornell, Ithaca, NY, United States, [email protected], Georgia Perakis
We consider an oligopoly market with more than two firms offering gross substitutes. We illustrate that while prices under Cournot competition remain higher than under Bertrand competition, the profit relationship is more complex and depends on the number of firms, their degree of market power, and quality differences among them. We examine several affine demand models and provide counterexamples where Bertrand profits can exceed Cournot profits along with an upper bound on the profit ratio.

I SC15
Procurement Auctions in OM
Cluster: Auctions and e-Commerce Invited Session
Chair: Wedad Elmaghraby, Assistant Professor, University of Maryland, Robert H. Smith School of Business, 4352 Van Munching Hall, College Park, MD, 20742, United States, [email protected] 1 - Supply Chain Coordination Under Information Acquisition Tunay Tunca, Professor, Stanford University, Graduate School of Business, 518 Memorial Way, Stanford, CA, 94305, United States, [email protected], Hyoduk Shin
We consider supply chain coordination with investment in demand forecasting. We show that common pricing schemes suffer from lack of coordination in production quantities and forecast. We show that this problem can be solved through flexible pricing. However, with private information, such mechanisms suffer implementation problems. We propose a contracting scheme that utilizes market interaction and resolves the implementation problems in addition to achieving full coordination.

2 - Estimating Choice Models from Airline Data Garrett Van Ryzin, Columbia University, Uris Hall, New York, NY, 10027, United States, [email protected], Gustavo Vulcano
We look at the results of an empirical study of estimating choice models from available airline data. We discuss both estimation methods and how the estimated choice models can be used to improve revenue management.

3 - Stochastic and Dynamic Price-Based Revenue Management with Nash Bargaining Nicola Secomandi, Assistant Professor, Carnegie Mellon University, Tepper School of Business, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected], Atul Bhandari
We consider the classical setting of a firm that sells a finite inventory of a single product within a finite number of periods. The seller negotiates the selling price with each buyer. We embed the Nash bargaining model in a dynamic and stochastic model where the seller dynamically optimizes the minimum acceptable price in each negotiation. We study properties of this model and show that its optimal value functions are higher than those obtained from the traditional posted-price model.

2 - Sole Sourcing vs. Second Sourcing: Impact of Demand Uncertainty Cuihong Li, Assistant Professor, University of Connecticut, 2100 Hillside Road, Unit 1041, Storrs, CT, 06269, United States, [email protected], Laurens Debo
We study the decision for a manufacturer (buyer) to select between sole sourcing and second sourcing mechanisms for a component with uncertain demand. Supplier capacity investment is required before the demand is realized. The incumbent supplier achieves cost reduction from learning. We find that a high or low capacity cost may both benefit second sourcing, and second sourcing is more favorable more when the buyer faces more demand uncertainty.

I SC17
Revenue Management Models with Customer Behavior
Contributed Session
Chair: Renato de Matta, Associate Professor, The University of Iowa, 108 PBB, Iowa City, IA, 52242, United States, [email protected] 1 - An Adaptive Dynamic Pricing Method for Self-Storage Industry Lijian Chen, Assistant Professor, University of Louisville, J.B. Speed School of Engineering, University of Louisville, Louisville, KY, 40292, United States, [email protected]
In the self-storage industry, the dynamic pricing is an important part for the department of revenue management. The most commonly used method is to adjust prices by the property’s performances, such as occupancy, competitor price, seasonal trend, and the unexpected events. We propose an alternative, which is a math programming based, adaptive method, to optimize the pricing policy for all the franchised properties. Our model provides a quantitative way to project the company wide revenue flow.

3 - Computational Analysis of Stable Coalition Structures in Sealed Bid Auctions Kemal Guler, Senior Scientist, Hewlett-Packard Labs, HP Labs MS1140, 1501 Page Mill, Palo Alto, CA, 94304, United States, [email protected], Bin Zhang
We study ex ante incentives to form coalitions in sealed-bid first-price and second-price auctions in private values environments. Auction subgames following a coalition formation stage involve asymmetries that necessitates a computational approach. We explore partitions of the set of bidders that are immune to deviations by individuals under exogenous sharing rules. We also explore the effects of risk aversion and dependence among bidder valuations on the stability of coalitions.

4 - Auction Design in the Presence of Non-Convex Costs Wedad Elmaghraby, Assistant Professor, University of Maryland, Robert H. Smith School of Business, 4352 Van Munching Hall, College Park, MD, 20742, United States, [email protected]
Auctions are increasingly becoming standard instruments for procurement. However, the performance of procurement auctions when suppliers have nonconvex costs are poorly understood. We present results from a series of experiments with human subjects at the University of Maryland.

2 - Decision Biases in Revenue Management: Some Behavioral Evidence J. Neil Bearden, Assistant Professor, University of Arizona, Eller College of Management, Mcclelland Hall 417, Tucson, AZ, 85720, United States, [email protected], Amnon Rapoport, Ryan Murphy
We study a simplified revenue management problem in which offers arrive sequentially and stochastically over a finite horizon. The value of each offer is random and revealed upon arrival, and the decision maker (DM) can accept fixed number of offers. Behavioral experiments reveal that DMs employ policies of the same form of the optimal policy, but show biases to demand too much in early periods and too little in later periods. We argue that behavioral work can compliment theoretical work in RM.

91

SC18

INFORMS PITTSBURGH — 2006 I SC19
Forestry Applications III
Sponsor: Energy, Natural Resources & The Environment Sponsored Session
Chair: Sandor Toth, Assistant Professor, Oklahoma State University, Stillwater, OK, 74078, United States, [email protected] 1 - Deciding Where to Establish Fuel Breaks on Flammable Forest Landscapes David Martell, Professor, Faculty of Forestry, University of Toronto, 33 Willcocks Street, Toronto, ON, M5S 3B3, Canada, [email protected], Wenbin Cui, Cristian Palma, Andres Weintraub
We partition a flammable forest landscape into cells and use a simple localized cell assessment procedure to rank cells with respect to their fire protection values. We then compare our results with the results obtained by using more exact models. Our cell ranking procedure can be used to help determine where to establish fuel breaks.

3 - Optimal Allocation of Flexible Products Under Consumer Choice Behavior Model Wenli Xiao, Fudan University, 59#501C, Lane 78, Wudong Road, Shanghai, 200433, China, [email protected]
Flexible product is a set of two or more alternatives. Suppliers could decide which alternative should be assigned to the customer. In reality, Consumers have product preferences. Considering the influence of consumer behavior in purchasing process, we provide a model to analyze the optimal allocation of flexible products. We use numerical simulations to represent that the increased demand for specific products once the flexible product was closed could result in higher revenue.

4 - Successive Approximation Based Controls for Revenue Management Problems Sreelata Jonnalagedda, Doctoral Student, University of TexasAustin, 3457 Lake Austin Boulevard E., Austin, TX, 78703, United States, [email protected]
We look at perishable asset revenue management problem in the context of an airline network. We use simulation and successive approximations to get closer to optimal solution. We compare bid price policy and certain equivalent policy to our approach, we provide a practical implementation direction for real world applications

5 - Dynamic Pricing in Reputation-Based E-Commerce Retailing Renato de Matta, Associate Professor, The University of Iowa, 108 PBB, Iowa City, IA, 52242, United States, [email protected], Dengfeng Zhang
We examine a revenue management problem involving a seller of perishable products in an e-commerce setting where buyers are sensitive to both price and seller reputation. Using a Markov decision process to model the problem over a finite horizon, we investigate the impact of changes in seller reputation and buyer heterogeneity on the seller’s optimal pricing strategy. The managerial intuitions behind our analytical and experimental results are presented.

2 - Strategic Budgeting for Wildfire Management in the U.S. Gyana Parija, IBM T.J. Watson Research Center, Yorktown Heights, NY, 10598, United States, [email protected], Tarun Kumar, Steve Carty
Field experiences on using an IP model and a solution approach for optimizing the initial response (IR) organization needed to maximize the effectiveness of various fire containment activities in the fire planning units (FPU’s) across the US are presented. These are large-scale IP problems requiring innovative solution approaches. Solution to the budgeting problem includes an IR organization that could consist of a set of hand crews, engines, bull-dozers, and helicopters.

I SC18
Advances in Geometric Programming
Cluster: Geometric Programming: Research Inspired by the Work of Dick Duffin, Elmor Peterson & Clarence Zener Invited Session
Chair: Jayant Rajgopal, University of Pittsburgh, Department of Industrial Engineering, 1039 Benedum Hall, Pittsburgh, PA, 15261, United States, [email protected] 1 - The Solution of General Finite Linear Systems via Flexible New GGP Dual Variational Principles Elmor Peterson, Systems Science Consulting, 3717 Williamsborough Court, Raleigh, NC, 27609, United States, [email protected]
These principles, based on “linearly constrained generalized GP duality”, provide infinitely many new “iterative methods” for solving each system of linear equations and/or inequalities, including every large-scale LP problem. Each method requires only the “unconstrained approximate minimization” of a convex function that can be “separable”. This “infinite flexibility” is likely to provide serious competition for previously developed iterative methodologies, such as “interior-point methodology”.

3 - Optimal Prevention and Detection of Invasive Species Robert Haight, Research Forester, USDA Forest Service, North Central Research Station, 1992 Folwell Avenue, St. Paul, MN, 55108, United States, [email protected], Shefali Mehta, Robert Venette
Non-native invasive species cause substantial environmental and economic damage to forests. Losses can be reduced by preventing arrival and rapidly detecting invasions. We include prevention and detection activities in a renewalreward process that accounts for uncertainty in pest arrival and detection. The objective is to allocate resources among prevention and detection activities to minimize expected cost of damages. Results are presented for oak wilt management in the Minnesota.

I SC20
The Value of Information
Sponsor: Decision Analysis Society Sponsored Session
Chair: Erin Baker, Assistant Professor, University of Massachusetts, Mechanical and Industrial Engineering, 220 Elab, Amherst, MA, 01003, United States, [email protected] 1 - The Value of Non-Definitive Experiments in Pure Science Erin Baker, Assistant Professor, University of Massachusetts, Mechanical and Industrial Engineering, 220 Elab, Amherst, MA, 01003, United States, [email protected]
Evaluating an experiment in the context of a decision problem is well understood. But this is not so well defined in a “pure science” situation, such as the goal of determining whether life ever existed on Mars. One can put a value on a definitive experiment, but many sequences of experiments result in probability distributions over the information of interest. We describe the difficulties in the situation, and propose methods for developing utility models over experiments.

2 - Geometric Programming Based Shape-Preserving Cubic L1 Splines Shu-Cherng Fang, Industrial Engineering and Operations Research, North Carolina State University, Raleigh, NC, 27695, United States, [email protected], Wei Zhang, John Lavery
Splines have been widely used for real world applications. One fundamental requirement for these applications is that they should have minimum nonphysical oscillations for data with arbitrary changes in magnitude and in node spacing. We present a geometric programming based cubic L1 splines to serve this purpose.

3 - Geometric Programming Operations Management Applications Carlton Scott, Professor, Paul Merage School of Business, University of California - Irvine, Irvine, CA, 92697, United States, [email protected], Thomas Jefferson
Geometric Programming evolved with a focus on cost optimization in engineeering design and many of the early applications (with few exceptions) were in this area. Another area where there is a wealth of application is Operations Management. In this talk, we overview previous literature in this area as well as giving several newer applications to production inventory control and project management.

2 - Information Acquisition and Technology Adoption Canan Ulu, PhD Candidate, Fuqua School of Business, Duke University, 1 Towerview Drive, Durham, NC, 27708, United States, [email protected], James E. Smith
We consider a technology adoption model where the firm is offered a technology whose benefits are uncertain. In each period, the firm can choose to adopt or reject the technology or gather information to reduce the uncertainty about the benefits of the technology. We generalize previous technology adoption models to allow more general distributions on benefits and to allow the benefits of the technology to be changing over time. We characterize the optimal value function and optimal policies.

92

INFORMS PITTSBURGH— 2006
3 - Value of Information is Systematically Biased Steve Greidinger, Policy Decisions Forum, PO Box 30374, Bethesda, MD, 20284, United States, [email protected]
The expected value of imperfect information determines whether an inaccurate test or prediction is cost-effective. Unfortunately, test results may not be properly interpreted or applied by decision-makers, and so we must discount the value of information in order to account for the possibility that the results will be ignored. Value of a test may fall precipitously given political roadblocks and decisionmaker indifference. Suggestions for avoiding underutilized studies will be supplied.

SC23

Daskin and Snyder (2005) to a two-echelon supply chain design problem. Our objective is to locate the primary facilities and then find the optimal reconfiguration of the supply chain when a disruption has occurred at a facility by assigning multiple backups.

3 - Applying Distribution Network Modeling to Improve Theater Distribution Operations Norman Reitter, Principal Logistics Engineer, Concurrent Technologies Corporation, 100 CTC Drive, Johnstown, PA, 15904, United States, [email protected], Jamie Baer
The study team used an iterative approach, applying network transhipment modeling, simulation modeling, and logistics analysis to create and test the best distribution network configuration with results shown in a distribution blueprint format. We provide an overview of the distribution planning problem and a decision support methodology used to provide distribution asset placement, fleet sizing, and process improvements to logistics operations.

4 - Strength of Preference Determines the Value of Information Philippe Delquié, Associate Professor of Decision Sciences, INSEAD, Boulevard de Constance, Fontainebleau, F-77300, France, [email protected]
Generalizing a previous result: The value of information about any and all alternatives of a decision situation is maximum when the decisionmaker is indifferent between the alternatives, and it vanishes as preference for one alternative gets stronger. This is true for any correlation structure among alternatives, utility function, and (random) wealth. This permits to define an “Intrinsic Value of Information” for any lottery: a measure of information value independent of the decision context.

4 - Logistics Composite Model Variance Reduction Alan Johnson, Session Chair, Air Force Institute of Technology, AFIT/ENS, 2950 Hobson Way, Building 641, W-P AFB, OH, 45433-7765, United States, [email protected], George Cole
The Air Force uses the Logistics Composite Model (LCOM) to determine aircraft maintenance manpower levels. We are investigating the extent to which variance reduction techniques can be successfully applied to this discrete event simulation, in the hope of improving the performance of existing LCOM optimizers. We discuss our progress and plans.

I SC21
Panel Discussion: Decision Analysis: Foundations in Mathematics, Science, and Engineering
Sponsor: Decision Analysis Society Sponsored Session
Chair: Ronald Howard, Professor and Director, Decisions and Ethics Center, Department of Management Science and Engineering, School of Engineering, Stanford University, Stanford, CA, 94305, United States, [email protected] 1 - Panel Discussion: Decision Analysis: Foundations in Mathematics, Science, and Engineering Moderator: Ronald Howard, Professor and Director, Decisions and Ethics Center, Department of Management Science and Engineering, School of Engineering, Stanford University, Stanford, CA, 94305, United States, [email protected], Panelists: Robert Winkler, James Matheson, Jim Dyer, Gordon Hazen, Ali Abbas
Decision analysis has contributed to decision-making in business, medicine, engineering, and law. Panelists discuss its origins, its relation to other disciplines, and the most appropriate academic preparation for a career in decision analysis. The panelists will first make brief presentations, then discuss among themselves, and finally invite audience participation.

I SC23
Joint Session New Frontiers/ED: Innovative OR Teaching in Industrial Engineering from the OR Division of IIE
Cluster: New Frontiers at the Intersection of Industrial Engineering and O.R. Invited Session
Chair: Sarah Ryan, Iowa State University, 2019 Black Engineering Bldg., Ames, IA, 50011-2164, United States, [email protected] 1 - Using Spreadsheets to Teach Operations Research and Industrial Engineering Raymond Hill, Associate Professor, Wright State University, 207 Russ Engineering Center, 3640 Col. Glenn Highway, Dayton, OH, 45435, United States, [email protected]
The modern spreadsheet has become arguably the standard analysis tool for business managers. As the analytical professionals, the Operations Research (OR)/ Industrial Engineering (IE) practitioners should also embrace the tools and capabilities of the spreadsheet. This talk covers the use of Solver for optimization applications, data analysis tools for statistical analysis and experimental design, and Visual Basic for Applications for decision support application development.

I SC22
Air Force Supply Chain Analysis
Sponsor: Military Applications Sponsored Session
Chair: Alan Johnson, Session Chair, Air Force Institute of Technology, AFIT/ENS, 2950 Hobson Way, Building 641, W-P AFB, OH, 454337765, United States, [email protected] 1 - Battle Damage Assessment (BDA) Enhancement Methodology Using Situational Parameter Adjustments Sharif Melouk, Assistant Professor of Operations Research, Air Force Institute of Technology, 2950 Hobson Way, Building 641, Wright-Patterson AFB, OH, 45433-7765, United States, [email protected], Marcus Perry, Michael Carras
In the context of close ground combat, the perception of BDA is closely linked with a soldier’s engagement decisions and has significant effects on the battlefield. This research examines the BDA process from a perception standpoint and proposes a methodology to collect pertinent data and model this perception. We propose a SME survey design and a method to model the BDA process as a discrete time Markov chain with Bayesian inference used to update probability distributions at each time step.

2 - Integrating O.R. Techniques Across the I.E. Curriculum Kim Needy, University of Pittsburgh, Department of Industrial Engineering, Pittsburgh, PA, 15261, United States, [email protected], Bryan A. Norman, Mary Besterfield-Sacre, Jayant Rajgopal
We present two main ideas to synthesize and integrate O.R. concepts across the undergraduate I.E. curriculum: (1) The explicit use of techniques learned in a traditional O.R. course in other I.E. classes in the curriculum, and (2) An unstructured problem solving course where students work in teams and are required to integrate their O.R. skills with other techniques to address the problem. We report on our observations based on two offerings of this class.

3 - A Global Supply Chain Project for International Student Teams Jo Min, Professor, Industrial and Manufacturing Systems Engineering, Iowa State University, 2019 Black, Ames, IA, 50011, United States, [email protected], Leopoldo Cardenas, Cheng-Kang Chen, John Jackman, Peter Ball
We describe the international student team project experience in modeling and analyzing a hypothetical global supply chain. The participants for this project include students at universities in Scotland, Taiwan, and Mexico as well as at the Iowa State University. Specifically, we show how the teams from the four nations are organized and managed. Also, we provide outcome assessment on various team aspects such as team dynamics and leadership. Lessons and future directions will be discussed.

2 - Reliable Supply Chain Design Considering Facility Failures Dinakar Gade, University of Arkansas, 4167 Bell Engineering Center, Department of Industrial Engineering, Fayetteville, AR, 72701, United States, [email protected], Edward Pohl
Modern Supply chains are complex and subject to disruption risks such as natural disasters or terrorist attacks. A supply chain risk profile depends on how its main components such as plants, warehouses, etc are configured. We extend

93

SC24

INFORMS PITTSBURGH — 2006 I SC25
Panel Discussion: OR Entrepreneurship #3
Cluster: OR Entrepreneurship Invited Session
Chair: Doug Samuelson, Homeland Security Institute, 8711 Chippendale Court, Annandale, VA, 22003, United States, [email protected] 1 - Panel Discussion: OR-Based Entrepreneurship Moderator: Doug Samuelson, Homeland Security Institute, 8711 Chippendale Court, Annandale, VA, 22003, United States, [email protected], Panelists: Joe Discenza, Deb Sadowski, Vijay Mehrota
While many members of INFORMS have worked in entrepreneurial firms, those who have actually started and run such companies are less numerous. These panelists, all of whom have succeeded as leaders of high-tech start-up companies, will discuss their experiences and lessons learned.

4 - Ill-Structured Problem Solving in an Optimization Course Sigurdur Olafsson, Assistant Professor, Iowa State University, Department of Industrial and Manufacturing Systems Engineering, Ames, IA, 50011, United States, [email protected]
Engineering problems are frequently ill-structured and many decision making problems are poorly defined with competing solutions. In optimization this occurs when assumptions and simplifications are needed for tractability. We describe an effort to bring ill-structured problem solving into a standard undergraduate optimization course. Using a web-based system, students identify the problem, make appropriate assumptions, and solve the simplified problem using standard optimization techniques.

I SC24
Micro-Institutional Processes
Sponsor: Organization Science Sponsored Session
Chair: Klaus Weber, Assistant Professor of Management and Organizations, Kellogg School of Management, Northwestern University, 2001 Sheridan Rd, Evanston, IL, 60208-2001, United States, [email protected] Co-Chair: Amit Nigam, Post Doctoral Researcher, Rotman School of Management, University of Toronto, 105 St. George Street, Toronto, ON, M5S 3E6, Canada, [email protected]
The papers in this session examine the linkages between institutional structures and local interactions and meaning making processes.

I SC26
Data Mining in Decision-Making
Sponsor: INFORMS Computing Society/Data Mining Sponsored Session
Chair: Victoria Chen, Associate Professor, University of Texas, Arlington, Industrial Engineering, Campus Box 19017, Arlington, TX, 76019-0017, United States, [email protected] 1 - Convex Version of Multivariate Adaptive Regression Splines for Optimization Dachuan Shih, PhD Candidate, University of Texas at Arlington, Dept. of Ind. & Manuf. Sys. Eng., Campus Box 19017, Arlington, TX76019, United States, [email protected], Victoria Chen, Seoung Bum Kim
Multivariate Adpative Regression Splines (MARS) is a flexible statistical modeling method that has been used in optimization for stochastic dynamic programs, Markov decision problems, and two-stage stochastic programming. Many optimization methods depend on convexity, but a nonconvex MARS approximation is inherently possible. To ensure MARS convexity: (1) The form of interaction terms is modified. (2) Coefficients are constrained to guarantee convex functions.

1 - Institutional Semiotics: Meaning and Markets Roy Suddaby, Pearson Faculty Fellow, School of Business, University of Alberta, 2-32E Business Building, Edmonton, AB, T7Y 1b(, Canada, [email protected]
This paper focuses on the rhetorical and semiotic strategies used by publicly traded companies to communicate to market actors; competitors, regulators, shareholdeers, employees and other market participants. We draw from rhetorical theory, semiotics and neo-institutional theory to demonstrate how organizations use public market disclosure mechanisms to code and deploy rhetorical “signals” simultaneously to multiple audiences.

2 - Micromobilization: Negotiating Institutional Change in Surgery through Cycles of Challenge in Workplace Encounters Kate Kellogg, Sloan School of Management, MIT, [email protected]
Institutional scholars present contradictory evidence regarding which actors in an organization are critical to effecting an actual rather than merely symbolic change in organizational work practices in response to coercive pressure. In this 15 month ethnography, I track the process of change in response to a reform involving surgical practices in two surgical teaching hospitals. In one of the hospitals, organization members accomplished a real change in practices. The change seems to have required, however, more than the actions of powerful top managers or middle status paraprofessionals, as current literature would suggest. Instead, real change in practices was accomplished via the mobilization of a cross-rank coalition. Members with non-traditionalist identities transformed themselves from a disorganized group of individuals into an organized cross rank coalition via interactions in everyday work encounters and used this cross-rank coalition to create pressure for change. The agentic processes which were critical to the change were 1) the building of a cross-rank reformer identity, frames and solidarity across ranks, 2) the use of the reformer identity and frames to marshal resources from top managers, 3) the pressuring of traditionalists via crisis, and 4) the cross-rank creation of a positive model for change.

2 - Platform Identification Using Data Mining for Product Family Design Seung Ki Moon, Graduate Research Assistant, The Pennsylvania State University, 233 Leonhard Building, University Park, PA, 16802, United States, [email protected], Timothy W. Simpson, Soundar R. T. Kumara
In this paper, we propose a methodology for identifying a platform along with variant and unique modules in a product family using fuzzy clustering. Fuzzy cmeans clustering is used to determine initial clusters based on the similarity among functional features. The clustering result is identified as the platform and the modules by a fuzzy membership function and classification. We apply the proposed methodology to determine a new platform using a case study involving a single-use camera family.

3 - Some Implications of Institutional Pluralism Matt Kraatz, College of Business, University of Illinois, Urbana-Champaign, [email protected], Emily Block
This paper elaborates some of the organizational implications of institutional pluralism. Institutional pluralism is the condition confronted by organizations that operate within more than one institutional sphere at the same time. Organizations in pluralistic contexts have multiple roles, obligations, and social identities which are bestowed upon them as a result of their multiple institutional affiliations. We discuss the challenges and opportunities that pluralism creates for the organization and also consider its implications for theory.

3 - Discriminant Analysis and Predictive Models in Medicine Eva Lee, Associate Professor, School of Industrial and Systems Engineering, Georgia Institute of Technology, Atlanta, GA, 30332, [email protected]
In this talk, we describe novel optimization-based predictive models, and analyze the process of selecting critical attributes through a feature-selection algorithm. Applications to prediction of cardiovascular disease, Alzheimer’s disease, human cancer, and the degree of cell membrane permeabilization via assisted ultrasound (for drug delivery) will be discussed.

4 - A Sequential Response Surface Methodology for the Aircraft Rotation Problem Siriwat Visoldilokpun, PhD Student, Department of Industrial and Manufacturing Systems Engineering, University of Texas at Arlington, PO Box 19017, Arlington, TX, 76019, United States, [email protected], Jay Rosenberger, Victoria Chen, H.W. Corley, Seoung Bum Kim
Most existing simulation optimization techniques are computationally intractable for large-scale problems. We develop a column generation approach for response surface methodology using multivariate adaptive regression splines (MARS). The algorithm is based upon Dantzig-Wolfe decomposition within Kelley’s cutting plane algorithm. We apply the approach to an aircraft rotation problem.

94

INFORMS PITTSBURGH— 2006 I SC27
Joint Session QSR/ Health Applications: Quality and Statistical Decision-Making in Healthcare Applications I
Sponsor: Quality, Statistics and Reliability, Health Applications Sponsored Session
Chair: Jing Li, Research Assistant, Department of Industrial and Operations Engineering, University of Michigan, 1205 Beal Avenue, Ann Arbor, MI, 48109, United States, [email protected] Co-Chair: Julie Simmons Ivy, Assistant Professor, University of Michigan, 701 Tappan Street, Ann Arbor, MI, 48109, United States, [email protected] Co-Chair: Nicoleta Serban, Assistant Professor, ISyE, Georgia Institute of Technology, 755 Ferst Drive NE, Atlanta, GA, 30332, [email protected] 1 - Statistical Process Control for Syndromic Surveillance Justin Chimka, Assistant Professor, University of Arkansas, Bell Engineering Center, Department of Industrial Engineering, Fayetteville, AR, 72701, United States, [email protected], Jia Zhou
The problem of syndromic surveillance will be considered here as one of analysis and interpretation of flu activity data collected and reported by the Influenza Branch at Centers for Disease Control and Prevention. We have measured specificity of individuals control charts for outbreak detection. Observations of isolates and influenza like illness have been analyzed alone and as functions of predictors including geographical region.

SC29

I SC28
Panel Discussion: 10 Years Anniversary for the QSR Section: Past, Current and Future
Sponsor: Quality, Statistics and Reliability Sponsored Session
Chair: Judy Jin, Associate Professor, Department of Industrial and Operations Engineering, University of Michigan, IOE 2855, Ann Arbor, MI, 48109, United States, [email protected] Co-Chair: Jye-Chyi (JC) Lu, Professor, Georgia Institute of Technology, 765 Ferst Drive, Atlanta, GA, 30022, United States, [email protected] 1 - Panel: 10 Years Anniversary for the QSR Section: Past, Current and Future Moderator: Judy Jin, Associate Professor, Department of Industrial and Operations Engineering, University of Michigan, IOE 2855, Ann Arbor, MI, 48109, United States, [email protected], Panelist: Jye-Chyi (JC) Lu
The panel is to review and access the QSR’s formation and growth in the past 10years, and to identify opportunities and define the future directions of the QSR. The panel will consist of 4~5 speakers, who are selected to represent the QSR organization leaders and three focused research areas leaders in Quality, Statistics, and Reliability respectively.

I SC29
Stochastic Models in Health Care
Cluster: Joint Cluster Healthcare/ HAS: Healthcare Engineering Invited Session
Chair: Elif Akcali, Assistant Professor, University of Florida, 303 Weil Hall, PO Box 116595, Gainesville, FL, 32611-6595, United States, [email protected] 1 - Patient Triage in the Aftermath of a Mass Casualty Event - A Dynamic Programming Approach Serhan Ziya, Assistant Professor, University of North Carolina at Chapel Hill, Dept. of Statistics and Oper. Research, 213 Smith Building, CB#3180, Chapel Hill, NC, 27599, United States, [email protected], Nilay Tanik Argon
We consider the patient prioritization (triage) problem that arises in the aftermath of mass casulaty events. Our objective is to identify effective triage decisions that maximize the expected number of lives saved, given the lifetime and treatment time distributions for patients and also the number of patients. We use stochastic dynamic programming to (partially) characterize the structure of the optimal policy. We propose a heuristic policy which is shown to work well by a numerical study.

2 - Automated Time Series Forecasting for Biosurveillance Galit Shmueli, Assistant Professor, University of Maryland, Robert H. Smith School of Business, Van Munching Hall, College Park, MD, 20742, United States, [email protected], Howard Burkom, Sean Murphy
For robust detection performance, alerting algorithms for biosurveillance require input data free of trends, day-of-week effects, and other systematic behavior. We describe three forecast methods and compare their predictive accuracy on each of 16 authentic syndromic data streams. The methods are 1) a nonadaptive loglinear regression model using a long historical baseline, 2) an adaptive regression model with a shorter, sliding baseline, and 3) the Holt-Winters exponential smoothing method.

3 - Nursing Home Care Quality: A Bayesian Network Approach Justin Goodson, Graduate Teaching Assistant, Department of Management Sciences, University of Iowa, S210 John Pappajohn Business Building, Iowa City, IA, 52242, United States, [email protected], Wooseung Jang
This study employs Bayesian networks to examine the multidimensional aspects of the quality of care in nursing homes. The findings indicate that nursing home quality is most accurately represented through a mix of structural, process, and outcome measures of quality. Furthermore, significant quality factors and their probabilistic relationships are identified.

4 - A Survey of the Statistical Methods on Health Care Fraud Detection Jing Li, Research Assistant, Department of Industrial and Operations Engineering, University of Michigan, 1205 Beal Avenue, Ann Arbor, MI, 48109, United States, [email protected], Kuei-Ying Huang, Judy Jin, Jianjun Shi
Frauds and abuses have incurred significant loss in the US health care expenditure. This paper conducts a comprehensive survey of existing statistical methods in health care fraud detection, focusing on the fraud characteristics, the uniqueness of health care data, the existing methods, and some discussions of opportunities for future R&D.

2 - A Strategic Model for Healthcare Resource Allocation Eylem Tekin, Assistant Professor, Texas A&M University, Industrial Engineering, College Station, TX, 77843, United States, [email protected]
We consider the problem of service capacity allocation to new patients in multiphysican healthcare facilities. Due to random patient arrivals, the allocation of workloads to physicians may not be always balanced, resulting in long waiting times. Taking into account that some patients would like to switch from their primary physican, we provide insights on the optimal allocation policy with the objective of minimizing long-term average costs.

3 - Optimizing Emergency Room Personnel Resources Chin-I Lin, Graduate Research Assistant, University of Florida, 303 Weil Hall, PO Box 116595, Gainesville, FL, 32611-6595, United States, [email protected], Murray Côté, Elif Akcali
We examine the personnel capacity planning problem for Emergency Room (ER) with two classes of patients and two types of health care teams, which have different processing capabilities and patient class - dependent service rates. We model the ER as a preemptive priority queuing system and develop approximations to compute the patient’s average system time. We then incorporate this into a network flow formulation to determine the optimal personnel capacity over a finite planning horizon.

95

SC30 I SC30
Pierskalla Award Presentations
Sponsor: Health Applications Section Sponsored Session

INFORMS PITTSBURGH — 2006
This paper deals with the containership stowage planning problem. The problem is stacking containers on different bays of a containership that visits several ports during a voyage. Due to its NP-Complete nature, exact algorithms for this problem are impractical. A heuristic approach based on Genetic Algorithm is proposed to deal with real world cases. Computational experiments and analysis of results as well as suggestions for future research are presented.

Chair: Julie Simmons Ivy, Assistant Professor, University of Michigan, 701 Tappan Street, Ann Arbor, MI, 48109, United States, [email protected] 1 - Optimization of Community Health Center Locations and Service Offerings with Statistical Needs Estimation Paul Griffin,Professor, Georgia Institute of Technology, 765 Ferst Drive, Atlanta GA 30332, United States, [email protected], Christina Scherrer, Julie Swann
Community Health Centers (CHCs) provide family-oriented health care services for people living in rural and urban medically under-served communities; they are an important part of the government’s plan to make health care more affordable. We develop an optimization model to determine the best location and number of new CHCs in a geographical network, as well as what specific services each CHC should offer at which level. We maximize the weighted demand coverage of the needy population subject to budget and capacity constraints, where costs are fixed and variable. We use statistical methods on national health databases to determine important predictors of healthcare need, and we apply these methods to Census data to obtain county-based estimates of demand. Using several healthcare performance metrics, we analyze the results of the system approach to location using the state of Georgia as a prototype. We apply our model to compare policies such as access versus coverage.

2 - Weight Annealing Heuristics for Solving Bin Packing Problems Edward Wasil, Professor of Management Science, American University, Kogod School of Business, Washington, DC, 20016, United States, [email protected], Bruce Golden, Kok-Hua Loh
We develop heuristics that use weight annealing to solve one-dimensional and two-dimensional bin packing problems. Our heuristics are easy to understand and contain a small number of parameters. We apply our heuristics to wellknown problem sets and find that they quickly produce high-quality solutions that are comparable to the best solutions found in the literature.

3 - Hybrid Evolutionary Algorithms for Rich Vehicle Routing Problems Andreas Reinholz, Chair of Systems Analysis, Department of Computer Science, University Dortmund, Schmelzer Weg, 8, Troisdorf, 53844, Germany, [email protected]
We present a methodology for designing Hybrid Evolutionary Algorithms for hierarchical nested Rich Vehicle Routing Problems (VRP). Problem specific neighborhood generating operators, working on different levels of the problem hierarchy, using problem special coding schemes, efficient data structures and accelerated function evaluations are introduced to develop competitive solvers. Experimental results with 101 new best solutions are given for benchmarks of several VRP.

2 - Improving Accessibility to Radiation Therapy for Cancer Patients Mariel Lavieri, Student, University of British Columbia, 2053 Main Mall, Vancouver, BC, V6T 1Z2, Canada, [email protected], Scott Tyldesley, Steven Shechter, John French, Tom Pickles, Martin Puterman
This talk describes a framework for modeling decision making for prostate cancer patients under constrained resource availability. We provide an overview of the proposed approach which will combine both clinical and operational decision making and then focus on patient specific decisions for regarding when to start radiation therapy.

4 - A Tabu Search Algorithm for the Single Vehicle Routing Allocation Problem John E. Beasley, Professor, CARISMA, Brunel University, Uxbridge, UB8 3PH, United Kingdom, [email protected]
The single vehicle routing allocation problem involves deciding a route for a vehicle (starting and ending at given locations) such that it visits some customers. Customers not visited can either be allocated to their nearest customer on the route, or can be isolated; where the objective is to minimise a weighted sum of routing, allocation and isolation costs. Our tabu search algorithm includes aspiration, path relinking and diversification. Computational results are presented.

3 - Statistical Surveillance and Process Control Methods in Health Care James Benneyan, Associate Professor, Northeastern University, 334 Snell Engineering Center, Boston, MA, 02115, United States, [email protected]
Statistical control charts have many important healthcare applications, including public health, epidemiology, physiologic monitoring, and care delivery processes. We discuss national healthcare measurement trends and the use and design of SPC methods for such cases. Technical considerations, recent contributions, and emerging opportunities are discussed, including risk-adjusted surveillance, weighted sequential probability ratio tests, and statistical deadband patient control schemes.

I SC32
Modeling in Financial Services
Sponsor: Financial Services Sponsored Session
Chair: Lyn Thomas, Professor of Management Science, School of Management, University of Southampton, Southampton, SO17 1 BJ, United Kingdom, [email protected] 1 - Procurement Strategies for Banks - Strategies for Optimizing Indirect Spend Procurement Lars Friedrich, Johann Wolfgang Goethe-University & E-Finance Lab, Breslauer Strasse 4b, Kelkheim, D-65779, Germany, [email protected]
Typically banks prefer “make” before “buy”. But this attitude has changed significantly due to explosive growth in Financial Services outsourcing. Compared to other industries, such as automobile, procurement in banking is still immature. This talk describes the current status of procurement and outsourcing in banking, develops a model on business dynamics of indirect spend procurement, describes strategies to optimize procurement with illustration using three empirical examples.

4 - To Put a Price on Life: Deriving a Meaningful Threshold for the Cost-Effectiveness of Medicine Chris Lee, Assistant Professor of Operations and Information Management, The Wharton School, University of Pennsylvania, 500 Jon M. Huntsman Hall, 3730 Walnut Street, Philadelphia, PA, 19104, United States, [email protected], Glenn Chertow, Stefanos Zenios
Cost-effectiveness analysis is often used to evaluate new medical technology for purposes of insurance coverage. The choice of appropriate threshold is critical but most thresholds that are frequently used are unjustifiable on theoretical grounds. We derive a threshold that appeals to the foundations of welfare economics. The method uses an optimization model and tries to uncover from practice patterns the price on life implicitly placed by the society.

I SC31
Metaheuristics and Vehicle Routing I
Sponsor: INFORMS Computing Society/Heuristic Search Sponsored Session
Chair: John E. Beasley, Professor, CARISMA, Brunel University, Uxbridge, UB8 3PH, United Kingdom, [email protected] 1 - A Genetic Algorithm for Container Loading Problem Ali Haghani, Professor and Chairman, Department of Civil and Environmental Engineering, University of Maryland, 1173 Glen L. Martin Hall, College Park, MD, 20742, United States, [email protected], Masoud Hamedi, Saini Yang

2 - How Much of the Gini is Due to Segmentation? Lyn Thomas, Professor of Management Science, School of Management, University of Southampton, Southampton, SO17 1 BJ, United Kingdom, [email protected]
Credit scoring systems are measured in many ways - predictive ability, power of discrimination, and accuracy of calibration. In practice a scoring system will consist of a number of scorecards - one for each subpopulation - and of vital importance is the overall ROC curve and Gini coefficient when the scorecards are combined. Part of this discrimination is due to choice of segments. This talk outlines an easily calculated bound on how much of the discrimination is due to the segmentation.

96

INFORMS PITTSBURGH— 2006
3 - Bayesian Inference for Rating Transition Models Catalina Stefanescu, Assistant Professor of Decision Sciences, London Business School, Regent’s Park, London, NW1 4SA, United Kingdom, [email protected], Radu Tunaru, Stuart Turnbull
We investigate stochastic models for rating transitions and propose a Bayesian framework for calibrating these models using Markov chain Monte Carlo techniques. The methodology is applied to the analysis of a recent data set of global corporate defaults and rating transitions from Standard and Poor’s, covering the period 1981 to 2004.

SC35

I SC34
Joint Session TSL/PPP: Operations Research and Non-Profit Applications
Sponsor: Transportation Science & Logistics, Public Programs and Processes Sponsored Session
Chair: Karen Smilowitz, Assistant Professor, Northwestern University, 2145 Sheridan Road, M233, EV3119, Evanston, IL, 60208, [email protected] 1 - OR/MS for Public-Sector Decisionmaking with Limited Resources: Values, Evidence and Methods Michael Johnson, Extern, Carnegie Mellon University, Center for Economic Development, 5000 Forbes Avenue, Pittsburgh, PA, 15213-3890, United States, [email protected]
Public decisionmaking is a challenging enterprise: practitioners and researchers must balance concerns of politics, policy, appropriate methods and fiscal constraints. In this presentation I review applications of OR/MS methods to problems in location of community corrections centers, fire stations and affordable housing, as well as reconfiguration of public school buildings and programs. My primary finding is that both qualitative and quantitative analytical methods are useful in assisting system change, especially in fiscally-constrained and politically contentious environments. Also, I find that “OR-like” thinking can be just as useful as explicit models and methods.

I SC33
Performance Analysis of the National Airspace System (NAS) Using Operations Forecasting
Sponsor: Aviation Applications Sponsored Session
Chair: Dipasis Bhadra, Principal Economist, MITRE/CAASD, 7515 Colshire Avenue, McLean, VA, 22102, United States, [email protected] 1 - Impact of Demand Forecast Uncertainty on National Airspace System Performance Estimates David Millner, The MITRE Corporation, 49185 Transmitter Road, Building 626, San Diego, CA, 92152, United States, [email protected], Jerome Freedman, Michael Callaham, Glenn Foster, William Baden
This study considers the effect of demand uncertainty on the benefits of future National Airspace System improvements. Historic variation is added to the FAA’s Terminal Area Forecast (TAF) to produce future demand ranges that preserve time correlations and airport dependencies. The Fratar algorithm is used to produce future demand and delays are estimated via simulation. The results show how demand ranges provide better information about future performance than could be derived by the TAF alone.

2 - Last Mile Distribution Problem for Humanitarian Relief Chains Burcu Balcik, Graduate Assistant, Department of Industrial Engineering, University of Washington, Box 352650, Seattle, WA, 98195, United States, [email protected], Benita Beamon, Karen Smilowitz
We consider last mile logistical operations in the humanitarian relief chain, which involves allocating relief supplies among demand locations from local distribution centers, and determining delivery schedules and routes for each vehicle throughout the planning horizon. We present an MIP model and solution approach for the last mile distribution problem. Finally, we present computational results and practical implications.

2 - Detailed Traffic Forecasting for the National Airspace System James Bonn, Research Analyst, The CNA Corporation, 4825 Mark Center Drive, Alexandria, VA, 22311, United States, [email protected]
The Federal Aviation Administration’s Air Traffic Organization’s Office of Strategy has been tasked with the generation of detailed future air traffic demand for the National Airspace System. We describe the process of turning Terminal Area Forecast and Enhanced Traffic Management System input data into the future forecasts using a Fratar technique. We also discuss predicted traffic levels at different Service Delivery Points.

3 - Sequential Allocation Problems for Non-Profit Agencies Robert Lien, Doctoral Student, Department of Industrial Engineering and Management Sciences, Northwestern University, 2145 Sheridan Road, Tech C210, Evanston, IL, 60208, United States, [email protected], Seyed Iravani, Karen Smilowitz
In this paper, we consider allocation decisions that arise in sequential inventory problems. In particular, we consider the Food Rescue Program operated by the Greater Chicago Food Depository.

3 - Probabilistic Forecasting at FAA’s Commercial Terminals Dipasis Bhadra, Principal Economist, MITRE/CAASD, 7515 Colshire Avenue, McLean, VA, 22102, United States, [email protected], Roger Schaufele
Generally, speaking, activities forecasts are point forecasts. This is true at the FAA as well as many other parts of the federal government. In this talk, we lay out a framework within which forecasts can be made probabilistic with uncertainty bands around them when the underlying model is unknown. Using Monte Carlo simulation, we apply this framework on total operations at FAA’s commercial terminals.

4 - Optimal Fundraising Categories Kevin McCardle, Professor, UCLA Anderson School, 110 Westwood Plaza, Los Angeles, CA, 90095-1481, United States, [email protected], Kumar Rajaram, Christopher Tang
We provide some data and a simple model for the setting of named categories in capital campaigns for charitable 501c3 organizations.

4 - FAA Future Demand Schedules for Air Traffic Planning John Gulding, Senior Strategist, Federal Aviation Administration, 800 Independence Avenue SW, Washington, DC, 20591, United States, [email protected]
Models used to assess future shortfalls must adapt a generic forecast to operational schedules that target key limitations in NAS performance. Schedule quality improves by reflecting seasonal variation and risk ranges around alternatives. For 2006, the ATO Office of Strategy developed a set of future schedules that allows for consistent planning among FAA program offices. The purpose of this paper is report on this development including core functions, verification and user implementation.

I SC35
Recent Development and Applications of the Continum Approximation Method
Sponsor: Transportation Science & Logistics Sponsored Session
Chair: Yanfeng Ouyang, Assistant Professor, University of Illinois at Urbana-Champaign, 205 N. Mathews Avenue, Urbana, IL, 61801, United States, [email protected] 1 - Design of a Distribution Network Anton Kleywegt, Associate Professor, Georgia Institute of Technology, School of Industrial Systems Engineering, Atlanta, GA, 30332, United States, [email protected], Jinpyo Lee, Amy Ward
We consider the design of a distribution network in which a less-than-truckload carrier transports shipments between many origins and destinations using several crossdocks. We describe a continuous approximation type approach that better captures the trade-off between the number of crossdocks that serves origins and destinations and the transportation distances than a widely used approach. We demonstrate our method with various instances.

97

SC36

INFORMS PITTSBURGH — 2006
photo storage. We find that when vendors face heterogeneous launching costs and different time lags in capturing profits (i.e., value latency), a vendor should compare its own launch cost and value latency with those of its rivals to decide its launch time.

2 - Modeling A Periodic Distribution Problem Peter Francis, PhD Candidate, Northwestern University, Department of IE/MS, 2145 Sheridan Road, Evanston, IL, 60208, United States, [email protected], Karen Smilowitz
We present a continuum approximation model for the Period Vehicle Routing Problem with Service Choice (PVRP-SC). The PVRP-SC is a variant of the Period Vehicle Routing Problem in which the visit frequency to nodes is a decision of the model. This variation can result in more efficient vehicle tours and/or greater service benefit to customers. The model helps evaluate the value of service choice in the strategic and tactical planning of periodic distribution systems.

3 - Is Consumer Demand Kinked? Estimating Menu Costs and Search Costs in Electronic Markets Anindya Ghose, Assistant Professor, Stern School of Business, New York University, 44 West Fourth Street, KMC 8-94, New York, NY, 10012, United States, [email protected], Bin Gu
The internet has significantly reduced consumers’ search costs and menu costs of retailers. But little is known about the magnitude of these costs in electronic markets. Using a dataset collected from Amazon.com and Barnes & Noble, we use product-level price and demand data to estimate these costs. We also identify the impact of search costs on market competition by exploring a unique insight of Stiglitz (1989) that search costs create kinkiness in aggregate demand when firms change prices.

3 - Discretizing Vehicle Routing Zones for Large-Scale Distribution Systems with Customer Uncertainty Yanfeng Ouyang, Assistant Professor, University of Illinois at Urbana-Champaign, 205 N. Mathews Avenue, Urbana, IL, 61801, United States, [email protected]
This paper proposes algorithms to automatically discretize vehicle routing zones (VRZ) from continuum approximation guidelines (i.e., shape and size requirements). VRZs are useful to the planning and operation of distribution systems in several practical contexts. We utilize a combination of spatial partitioning techniques to systematically obtain optimum zone designs that conform to the guidelines.

I SC38
RASIG Roundtable I - Railroad Capacity Network Allocation and Investment
Sponsor: Railroad Applications Sponsored Session
Chair: Carl Van Dyke, Director, Mercer Management Consulting, 125 Village Boulevard, Suite 270, Princeton, NJ, 08540, United States, [email protected] 1 - Impact of Capacity Constraints on Operating Plan Design Alan Blumenfeld, CSX Corporation, Inc., 500 Water Street, Jacksonville, FL, 32202, United States, [email protected]
Developing an operating plan where your network has significant areas of capacity constraints requires a different approach. This approach requires the full integration of business processes, Plan Development, Yield Management, and Tactical operations. This discussion will address how this full integration can help produce a more fluid and profitable operation.

I SC36
Panel Discussion: Integrating Demand and Network Models
Sponsor: Transportation Science & Logistics Sponsored Session
Chair: Joan Walker, Boston University, 675 Commonwealth Avenue, Boston, MA, 02215, United States, [email protected] 1 - Panel Discussion: Integrating Demand and Network Models Moderator: Joan Walker, Boston University, 675 Commonwealth Avenue, Boston, MA, 02215, United States. Panelists: Moshe Ben-Akiva, Hani Mahmassani, Srinivas Peeta, Peter Vovsha
In research and practice, there have been many advances on both the travel demand modeling (e.g., activity-based models) and travel supply modeling (e.g., traffic microsimulation and dynamic traffic assignment). However, there has been less attention to the integration of these advanced models. Current integration techniques (e.g., loose coupling of these models to transfer data from one to the other) do not take advantage of the added behavioral data available in each model, nor do they achieve consistency among parameters that are used in both demand and supply models. This panel session brings together top researchers with experience in both demand- and supply-side modeling to discuss the status of the field, critical issues, innovative research, and future directions.

2 - Use of Operations Research Models to Plan Capacity at Norfolk Southern Wayne Mason, Senior Director of Strategic Planning, Norfolk Southern Corporation, 1200 Peachtree Street NE, Atlanta, GA, 30309, United States, [email protected]
Norfolk Southern and other North American railroads have been experiencing unprecedented growth in recent years. Facing the challenges of a growth industry, Norfolk Southern turned to Operations Research models and techniques to draw more capacity out of the existing operating plan and infrastructure. Furthermore, OR tools are being employed to identify the “choke points” in the NS transportation network and make recommendations to invest in new capacity.

I SC37
Economics of IS
Sponsor: Information Systems Society Sponsored Session
Chair: Vidyanand Choudhary, Assistant Professor, University of California, Irvine, Paul Merage School of Business, SB 313, Irvine, CA, 92697, United States, [email protected] 1 - A Model to Compare an Online Intermediary Strategy versus an Online Direct Retailer Strategy Fred Riggins, Assistant Professor, University of Minnesota, Carlson School of Management, 321 19th Avenue S., Minneapolis, MN, 55455, United States, [email protected]
We develop a model to compare an online direct retailer with an online intermediary. In the absence of transaction cost advantages the direct retailer strategy dominates due to a supply dependency for the intermediary. However, the intermediary strategy may be better due to lower entry costs. The best strategy may be to enter the market as an online intermediary and evolve to become a direct retailer. We model how storefront technology maturity and adoption of e-commerce impacts this evolution.

3 - The Rail Network Capacity Crunch: Public and Private Goals David Hunt, Cambridge Systematics, Inc., PO Box 816, Pennington, NJ, 08534-0816, United States, [email protected], Raphael Kedar
Freight volumes are at record levels, creating capacity issues for all transportation modes. The responsibility for expanding capacity on roads, ports, and airports falls on the public sector, while railroad capacity is a private sector responsibility. There is often a disconnect between public and private goals. This presentation looks at the capacity crunch and different visions for the rail network. It also explores operations research methods needed to advance public and private goals.

4 - Dynamic Planning and Implementation: Managing Emergency Rail Outages Trent Sommers, System Manager Network Optimization, Norfolk Southern Corporation, 1200 Peachtree Street, NE, Box 158, Atlanta, GA, 30309, United States, [email protected]
An overview of Norfolk Southern’s service design philosophy and systems structures. How NS utilizes its integrated transportation data systems for capacity management during rail outages will be explored, using Hurricne Katrina as an example. A look at how implementation capabilities and short time horizons impact rail transportation planning and execution strategies. Consideration of change management process during systems development will also be explored.

2 - To Preempt or Not to Preempt? A Timing Decision Model for Software Functionality Additions Amy Ping Wu, Carlson School of Management, University of Minnesota, Information and Decision Sciences, 321 Nineteenth Avenue South, Suite 3-365, Minneapolis, MN, 55455, United States, [email protected], Robert Kauffman
We present a game-theoretic model to study the optimal timing for vendors to add new software functionality to their technology products. An example of such functionality additions is the expansion of portals’ capabilities to provide digital

98

INFORMS PITTSBURGH— 2006 I SC39
Where Do We Want to Go in Education?
Cluster: Where Do We Want to Go?
(In observance of Arthur Geoffrion’s retirement)

SC41

3 - Solution Approaches for Facility Location of Medical Supplies for Large-Scale Emergencies Hongzhong Jia, Research Assistant, University of Southern California, 2113 Oak Street, Los Angeles, CA, 90007, United States, [email protected], Fernando Ordonez, Maged Dessouky
We propose models and solution approaches for locating the medical supply facilities in response to large-scale emergencies. We address the problem’s uncertainty by ensuring each demand point is serviced by multiple facilities at different distances. Three heuristics are developed to solve this problem: genetic algorithm, locate-allocate, and Lagrangean relaxation heuristics. We evaluate the performance of the model and the heuristics on illustrative large-scale emergency examples.

Invited Session
Chair: Christopher Tang, Professor, Anderson School, University of California - Los Angeles, 110 Westwood Plaza, Los Angeles, CA, 90095, United States, [email protected] 1 - Engineering Education and Operations Research Thomas Magnanti, Institute Professor and Dean of Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, 1-206, Cambridge, MA, 02139, United States, [email protected]
Considerable national dialog during the past few years has focused on the perceived need for reforms in both the content and delivery of higher education. In particular, as reflected by the National Academy of Engineering’s 2020 report, considerable attention has focused on engineering education. This talk will highlight some of these deliberations as well as their potential implications for education in operations research.

4 - Cannibalization in a Competitive Environment Tammy Drezner, Professor, California State University, Department of ISDS, Fullerton, CA, 92834, United States, [email protected]
When a retail chain is expanding by adding outlets, two incompatible objectives have to be adressed: 1) The total market share captured by the chain should be maximized; 2) Cannibalization of chain members’ market share should be minimized. The models are formulated and analyzed.

2 - Where Do We Want OR to Go in Education and Vice Versa? John R. Birge, Professor, University of Chicago, Graduate School of Business, 5807 S. Woodlawn Avenue, Chicago, IL, 60637, United States, [email protected]
While academics still represent the largest group of operations researchers judged by INFORMS membership and meetings attendance, education remains ironically substantially untouched by OR. Perhaps not so coincidentally, OR education also remains relatively unchanged by recent developments in education. In this talk, I will discuss possible channels for OR both to have a greater role in education developments and to benefit more from innovations in education.

I SC41
Diffusion Approximations and Applications
Sponsor: Applied Probability Sponsored Session
Chair: Sunil Kumar, Professor of Operations, Information, and Technology, Stanford University, Graduate School of Business, 518 Memorial Way, Stanford, CA, 94305-5015, United States, [email protected] 1 - Long Time Asymptotics for Constrained Diffusions in Polyhedral Domains Amarjit Budhiraja, Professor, University of North Carolina, Statistics and Operations Research, CB # 3260, Chapel Hill, NC, 27599, United States, [email protected], Chihoon Lee
Long time asymptotics of constrained diffusions arising in the heavy traffic analysis of queueing networks is studied. We strengthen the result of Dupuis and Williams(1994) that showed ergodicity of an SRBM by establishing geometric ergodicity. Geometric ergodicity is also obtained for a broad family of constrained diffusions with state dependent coefficients. As consequences, we obtain finite exponential moments of the invariant measure, uniform time estimates on moments and FCLTs.

3 - What Skills Does Industry Want from Our Graduates? ManMohan Sodhi, Professor, Cass Business School, City University London, 106 Bunhill Row, London, EC1Y8TZ, United Kingdom, [email protected]
We describe the skills that industry employers seek from graduates of OR programmes with a view to motivate discussion on what should be taught and how it should be taught in OR or OM programmes. Our results are based on content analysis of detailed job ads. We also discuss what industry (implicitly) defines OR and OM to be and therefore the possible implications on education.

I SC40
Medical Location Issues
Sponsor: Location Analysis Sponsored Session
Chair: Tammy Drezner, Professor, California State University, Department of ISDS, Fullerton, CA, 92834, United States, [email protected] 1 - A Location Model of an Eyeglass Lens Adjustment Problem Richard Francis, Professor Emeritus, Department of Industrial & Systems Engineering, University of Florida, 303 Weil Hall, Gainesville, FL, 32611, United States, [email protected]
The problem is to find one (or several) best eyeglass lens adjustment(s) for an eye with double vision when the vision changes throughout the day. We model the problem as a well-known location problem, and mention some personal experience. Known theoretical results have some interesting applications, believed to be new.

2 - Critically Loaded Queueing Systems that Behave as Underloaded Rami Atar, Associate Professor, Technion - Israel Institute of Technology, Department of Electrical Engineering, Haifa, 32000, Israel, [email protected], Avi Mandelbaum, Gennady Shaikhet
We consider a multiclass queueing system with a fixed number of classes and a fixed number of service stations, in the Halfin-Whitt diffusion asymptotic regime. The underlying critically loaded fluid model may or may not be “throughput optimal” in a sense that can be made precise. We discover a new type of “heavy traffic behavior” where it is possible to maintain empty queues with probability approaching one. We show that this can be achieved if and only if throughput optimality does not hold.

2 - Incorporating Congestion in Designing Preventive Healthcare Facility Networks Yue Zhang, PhD Candidate, McGill University, 1001 Sherbrooke Street West, Montreal, QC, H3A 1G5, Canada, [email protected], Oded Berman, Vedat Verter
The problem of locating an undetermined number of preventive healthcare facilities to maximize participation is proposed. Each facility is an M/M/1 queue, and each customer patronizes the facility with the minimal expected total time (travel time plus the expected time customers spend at the facility). Facilities must achieve a minimum workload requirement. Four heuristics are proposed and compared. A real life application (breast cancer screening centers in Montreal) is examined and analyzed.

3 - A Dual Approach to Singular Stochastic Control Problems Sunil Kumar, Professor of Operations, Information, and Technology, Stanford University, Graduate School of Business, 518 Memorial Way, Stanford, CA, 94305-5015, United States, [email protected], Joy Rajiv
Singular stochastic control problems arise in a variety of applications ranging from economics to finance to queueing systems. Except in special cases, these problems can only be solved numerically. We outline a numerical method based on duality for solving such problems. We iteratively construct dual feasible solutions, by modifying the cost structure, that approach the primal optimal solution. We illustrate the method on two well-known problems.

4 - Asymptotic Optimality of a Fixed-Point-Based Threshold Control David D. Yao, Professor, Department of Industrial Engineering and Operations Research, Columbia University, New York, NY, 10027, [email protected], Heng-Qing Ye
In a prior work, Li and Yao, a stochastic network with simultaneous resource occupancy is studied, and a threshold control policy is proposed based on a fixedpoint approximation. Here, we establish the asymptotic optimality of this control policy under fluid and diffusion scaling, and discuss its application in airline revenue management.

99

SC42 I SC42

INFORMS PITTSBURGH — 2006
2 - The First Ten Years of Using The Analytic Network Process (ANP): A Literature Review Enrique Mu, Director, MIS Program, University of Pittsburgh, 345 Mervis Hall, Pittsburgh, PA, 15260, United States, [email protected]
Since its formal proposal by Saaty, ten years ago, the ANP has had only a few applications and has appeared rather sparsely in the literature. Still, the body of ANP articles is growing and at this point merits careful examination. This paper will review the current literature on ANP with an emphasis on its applications.

Improving Scheduling Decision-Making
Cluster: Scheduling Invited Session
Chair: Jeffrey Herrmann, Professor, University of Maryland, Department of Mechanical Engineering, College Park, MD, 20742, United States, [email protected] 1 - Heuristics for Scheduling a Single Machine with FamilyDependent Setup Times Reha Uzsoy, Professor, Purdue University, 315 N. Grant Street, West Lafayette, IN, 47907, United States, [email protected], Juan Diego Velasquez
We examine the problem of minimizing maximum lateness on a single machine with family dependent setup times. We develop several heuristics including a rolling horizon approach and an incomplete dynamic program based on the algorithm of Monma and Potts(1989). Computational results are provided.

3 - Business Internal Audit Prioritization: A Case Study Wen Chiang, Professor, University of Tulsa, 118-J BAH, 600 S. College Avenue, Tulsa, OK, 74104, United States, [email protected], Jennifer Shang
Internal audits evaluate internal control of a business through financial, operational, and compliance review. It assesses the risk of asset loss, studies business processes, and identifies opportunities to improve efficiency and effectiveness. In this paper, we apply AHP and DEA to determine the business units that need audit. The framework is comprehensive, flexible, and shows great potential for internal audit prioritization and resource allocation.

2 - Scheduling Final Exams at Cornell University Dmitriy Levchenkov, PhD Student, Cornell University, School of OR&IE, 295 Rhodes Hall, Ithaca, NY, 14853, United States, [email protected], Robert Bland, David Shmoys
Final exams at Cornell are scheduled in 21 periods on 7 days, by first grouping all courses into 1 of 14 groups based on their class time or, for courses with multiple sections, in one of 6 exception groups. Student complaints about taking 3 exams in a day or 2 consecutive exams prompted us to develop an IP approach to the problem. This approach has been adopted here with great success. Somewhat surprisingly, the solution of the required IPs has not been an obstacle in this work at all.

I SC44
JFIG Paper Competition Finalists I
Sponsor: Junior Faculty Interest Group (JFIG) Sponsored Session
Chair: Alan Scheller-Wolf, Associate Professor, Tepper School of Business, Carnegie Mellon University, Pittsburgh, PA, 15213, United States, [email protected] 1 - JFIG Paper Competition Finalists
In these two sessions the finalists of the 2006 JFIG paper competition will present their research. The paper competition is organized by the Junior Faculty Interest Group (JFIG) forum in INFORMS. JFIG was created in 2001 to promote the career development of tenure-track faculty in INFORMS. The goals of the paper competition are to encourage research among junior faculty and to increase the visibility of research conducted by junior faculty within the fields of operations research and management science. Six finalists are selected by the award committee and their papers are presented in these two sessions. The papers submitted for the competition are evaluated based on the importance of the topic, appropriateness of the research approach, and the significance of research contribution. Winners are announced at the JFIG lunch on Sunday at the INFORMS Annual Meeting.

3 - Integrating Organizational, Decision-Making, and ProblemSolving Perspectives of Scheduling Jeffrey Herrmann, Professor, University of Maryland, Department of Mechanical Engineering, College Park, MD, 20742, United States, [email protected]
Production scheduling activities are common but complex. This leads to many different perspectives. Each has a particular scope, its own set of assumptions, and a different approach to improving production scheduling. Three important perspectives are the problem-solving perspective, the decision-making perspective, and the organizational perspective. This talk presents an integrative strategy for selecting, in a particular setting, an approach for improving production scheduling.

4 - Scheduling Reentrant Jobs Ed Mooney, Associate Professor, Montana State University, M&IE Department, Bozeman, MT, 59717-3800, United States, [email protected], Deepu Philip
Capacitated workstations, multiple routes, sequence dependent setup times, and reentrant jobs make the reentrant flexible job shop with sequence-dependent setups (RFJSSDS) problem one of the most general and difficult shop scheduling problems. We identify several important properties of RFJSSDS and use them to couple a local search algorithm with a general shop simulator for efficient scheduling. Experiments with a set of synthetic instances demonstrate the value of the approach.

I SC45
Tutorial: Model Uncertainty, Robust Optimization and Learning
Cluster: Tutorials Invited Session
1 - Model Uncertainty, Robust Optimization and Learning Andew Lim, Professor, University of California, Department of IEOR, Berkeley, CA, 94720, United States, [email protected], Z.-J. Max Shen, George Shanthikumar
We discuss different models of model uncertainty with variable, parameter and measure uncertainty sets, and present different approaches to robust optimization with and without bench-marking. Two alternative learning approaches, Objective Bayesian Learning and Operational Learning will be discussed. Applications considered for illustration are the classical inventory control problem, the inventory control problem with censored demand data and the portfolio selection problem.

I SC43
AHP/ANP Applications in Operations Management
Cluster: Analytic Hierarchy Process Invited Session
Chair: Jennifer Shang, Associate Professor, University of Pittsburgh, 254 Mervis Hall, Roberto Clemente Drive, Pittsburgh, PA, 15260, United States, [email protected] 1 - Offshore Outsourcing Decision Making: A Government Policy Focus Tammy Tjader, PhD Candidate, University of Pittsburgh, 343 Mervis Hall, Pittsburgh, PA, 15260, United States, [email protected]
The need for US firms to stay competitive along with the job loss distress has kindled fierce debate about offshore outsourcing. To provide a satisfactory government guideline for offshore outsourcing, we investigate the pertinent factors and propose a comprehensive evaluation framework to address this sensitive and important issue.

I SC46
Tutorial: Enhance Your Own Research Productivity Using Spreadsheets (From Sin to Salvation)
Cluster: Tutorials Invited Session
1 - Enhance Your Own Research Productivity Using Spreadsheets (From Sin to Salvation) Janet Wagner, Associate Professor, University of MassachusettsBoston, MSIS Department, 100 Morrissey Boulevard, Boston, MA, 02125, United States, [email protected], Jeffrey Keisler
Spreadsheets are preferred by many OR/MS modelers. However, user surveys indicate that most do not use basic good practices and most large spreadsheets are seriously flawed. Fortunately, solutions to many of these problems are widely

100

INFORMS PITTSBURGH— 2006
available and easy to learn. The authors, who have first hand experience with both “sin” and “salvation” in the spreadsheet kingdom, present “top ten” Excel methods and four major applications reflecting personal and professional experiences. Bring your laptops! impact of additional suppliers.

SC49

I SC47
New Supply Chain Initiatives
Cluster: Supply Chain and Operations Engineering Invited Session
Chair: Zuo-Jun Max Shen, Professor, University of California at Berkeley, 4141 Etcheverry Hall, University of California, Berkeley, CA, 94720, United States, [email protected] 1 - Supply Uncertainty and the Reverse Bullwhip Effect Lawrence Snyder, Assistant Professor, Department of Industrial & Systems Engineering, Lehigh University, 200 West Packer Avenue, Bethlehem, PA, 18015, United States, [email protected], Zuo-Jun Max Shen, Ying Rong
We present theoretical and empirical evidence for a reverse bullwhip effect (RBWE) that occurs in a variety of situations involving supply uncertainty. In the RBWE, order volatility increases as one moves downstream in the supply chain. This is the opposite of the classical bullwhip effect, which is generally associated with demand uncertainty. We examine causes of RBWE, discuss its impact, and suggest strategies for mitigating it.

2 - Inducing Retailer Forecasting: Rebates versus Returns Terry Taylor, Associate Professor, Columbia University, Graduate School of Business, 3022 Broadway, Uris 417, New York, NY, 10027, United States, [email protected], Wenqiang Xiao
We examine how a retailer’s ability to improve her demand information by exerting forecasting effort affects what type of contract a manufacturer should offer. We show whether a manufacturer should offer a contract that rewards the retailer for selling units (rebates) or a contract that rewards the retailer for not selling units (returns).

3 - Optimal Design of Supply Contracts Under Information Asymmetry Ying-Ju Chen, Doctoral Candidate, Stern School of Business, New York University, 44 W. 4th Street KMC 8-152, New York, NY, 10012, United States, [email protected], Sridhar Seshadri
We consider a supply chain where multiple suppliers with heterogeneous production costs and salvage-values sell through a single retailer. The suppliers’ cost structures are either independent, private but correlated, or completely known amongst themselves. We investigate the optimal procurement mechanism from the retailer’s perspective under different scenarios of information asymmetry.

I SC49
Alan S. Manne (1925-2005): Innovative Energy, Environment and Operations Modeler
Cluster: History of OR and Management Science Invited Session
Chair: Arthur F. Veinott, Jr., Professor, Department of Management Science & Engineering, Stanford University, 309 Terman Engineering B, Stanford, CA, 94305, United States, [email protected] 1 - Alan S. Manne: Professional Life and Early Creative Operations Modeling Arthur F. Veinott, Jr., Professor, Department of Management Science & Engineering, Stanford University, 309 Terman Engineering B, Stanford, CA, 94305, United States, [email protected]
Alan was precocious, graduating from Harvard at age 18. His Ph.D. thesis, Scheduling of Petroleum Refinery Operations, influenced the oil industry’s adoption of linear programming in the 1950s. In 1958 he showed how to approximate notoriously difficult multiproduct dynamic economic-lot-size models with shared capacity by a linear program. In 1960 he introduced an ingenious linear-programming approach to Markov decision chains. We review these works, his professional career and his many honors.

2 - Discrete Location with Discrete Choice Chung-Piaw Teo, Associate Professor, National University of Singapore, Business School, 21 Lower Kent Ridge Road, Singapore, 119077, Singapore, [email protected]
Consider a discrete facility location problem where each consumer has a random utility attached to each facility. Given a set of opened facilities, the consumer will choose which facility she wants to be served in. We would like to find the optimal facility location solution to maximize social welfare, after taking into account the facility setup cost. We use a new discrete choice model proposed recently by Karthik, Song and Teo to reduce this problem to a simple convex IP.

3 - An Application of Operational Statistics Leon Chu, Assistant Professor, University of Southern California, Marshall School of Business, Los Angeles, CA, 90089, United States, [email protected], George Shanthikumar, Zuo-Jun Max Shen
Operational statistics, an approach that integrates parameter estimation and optimization, leads to better solutions comparing with the traditional approach. This can be verified analytically for some simple problems. In this talk, we illustrate how to apply operational statistics to more complex problems.

4 - Coordinating Efforts of Multiple Retailers in a Decentralized Supply Chain Stephen Shum, PhD Candidate, Massachusetts Institute of Technology, 77 Massachusetts Avenue, E40-130, Cambridge, MA, 02139, United States, [email protected], David Simchi-Levi
We study the impact of effort in a supply chain with multiple retailers. The costly effort engaged by a retailer may increase or decrease the demands of other retailers. We show that traditional contracts coordinate only under very restrictive conditions. We propose revenue sharing with fixed target rebate. These contracts are shown to be both coordinating and flexible. Interestingly, similar contracts have been used in the fresh food industry to motivate retailers to promote the products.

2 - Alan S. Manne: Early Contributions to Sectoral and EconomyWide Planning Models Donald Erlenkotter, Professor Emeritus, Anderson Graduate School of Management, University of California, Los Angeles, CA, United States, [email protected]
During the 1960s and early 1970s, Alan Manne made fundamental contributions in the development and application of sectoral and economy-wide planning models. These included models for capacity expansion and location, and interindustry models for planning economic development. His work is exemplary in its meticulous attention to data and its interplay between theory-based applications and applications-grounded theory. We review his contributions to these areas and their impact on the field.

I SC48
Supply Chain Management with Asymetric Information
Sponsor: Manufacturing & Service Operations Management Sponsored Session
Chair: Izak Duenyas, Professor, Ross School of Business, 701 Tappan Street, Ann Arbor, MI, United States, [email protected] 1 - Sealed-Bid vs. Open Total-Cost Procurement Auctions: Should the Buyer Reveal Suppliers’ Quality? Dimitris Kostamis, PhD Candidate, University of Michigan, 1205 Beal Avenue, Ann Arbor, MI, 48109, United States, [email protected], Damian Beil, Izak Duenyas
We compare two auctions for procurement. A supplier’s effective bid is a function of their bid price and an adjustment assigned by the buyer, capturing factors such as expected quality. Each supplier’s adjustment is private information to him and the buyer. The buyer can (implicitly) announce her private adjustment cost information to all suppliers via an open auction, or conceal it by running a sealed-bid auction. We characterize the buyer’s choice, and study the

3 - Alan S. Manne: Energy Modeling Pioneer John P. Weyant, Professor, Department of Management Science and Engineering, Stanford University, Terman Engineering B, Stanford, CA, 94305, United States, [email protected]
The late 1960s and early 1970s marked the genesis of modern energy modeling. New problems were addressed, new data was collected, new algorithms were implemented, and new insights were developed. In all these areas Alan S. Manne was a trail blazer. This talk highlights Alan’s contributions using early energy market analyses and energy technology assessments as examples.

4 - Alan S. Manne: Energy Vulnerability Hung Po Chao, Director of Market Monitoring, Independent Systems Operator, New England, One Sullivan Road, Holyoke, MA, 01040, United States, [email protected]
Alan Manne made many fundamental contributions to energy policy issues. His work characteristically offers simple and yet profound insights. This talk highlights his prescient insights of energy vulnerability in a modern economy characterized by complex interactions, interdependence and uncertainties, a common theme that can be found in many of his path-breaking works.

101

SC50

INFORMS PITTSBURGH — 2006 I SC51
Applications of Data Mining/Machine Learning
Contributed Session
Chair: Bin Zhang, IBM China Research Laboratory, Building 19 Zhongguancun Software Park, Beijing, China, [email protected] 1 - Operations Research and Data Mining as Complementary Modeling Approaches Stephan Meisel, Research Associate, Technical University Braunschweig, Abt-Jerusalem-Str. 4, Braunschweig, Germany, [email protected]
Purpose of our work is to demonstrate that Operations Research and Data Mining are complementary approaches for the single task of modeling complex real-world operations. An integrative view on the two fields is developed on a conceptual level. Problems that cannot be treated adequately by use of strict mathematical modeling approaches only (e.g. Dynamic Vehicle Routing Problems) are supposed to be tackled according to our framework.

5 - Alan S. Manne: Innovative Approaches to Climate Policy Design Thomas Rutherford, 4829 Northgate Drive, Ann Arbor, MI, 48103, United States, [email protected]
Alan Manne was a pioneer in the adaptation of energy-economy models to climate policy issues. The talk will highlight aspects of these models which were pathbreaking and remain state of the art: putty-clay technology, stochastic equilibria, integrated treatments of climate and economic systems, and the solution of large-scale equilibrium models with tax distortions and public goods solved by use of iterative optimization.

I SC50
Network Design
Sponsor: Telecommunications Sponsored Session
Chair: Abdullah Konak, Assistant Professor, Penn State - Berks, Tulpehocken Road, Reading, PA, 19610, United States, [email protected] 1 - Models and Heuristics for the Node-Degree Constrained Minimum Spanning Tree Problem Christophe Duhamel, Universite Blaise Pascal, LIMOS, ISIMA, Campus des Cézeaux, BP 10125, AubiËre Cedex, 63173, France, [email protected], Luis Gouveia, Pedro Moura, Mauricio Souza
The NDCMST problem consists in finding a minimal cost spanning tree given an upper limit on the degree of each vertex in the tree. A concave node degree cost function is added to the objective function. Starting from a valid formulation, we propose a reformulation by discretization. Then we present a lagrangean relaxation and another reformulation with the same lower bound. Upper bounds are computed using a hybrid GRASP/VND heuristic. Computational results are shown on medium-sized instances.

2 - Product Data Mining as a Tool for Improving Productivity Pamela Ajoku, Research Assistant Professor, University of Pittsburgh, Industrial Engineering, 1048 Benedum Hall, Pittsburgh, PA, 15260, United States, [email protected], Bart Nnaji
Manufacturing environments have begun applying data mining methodologies as a result of the vast amounts of data that can be generated within a single plant on a daily basis. Often multiple plants and collaborative partners are connected, compounding data analysis and knowledge discovery on a national and global level. Data mining applications within the manufacturing environment include market analysis, product design, manufacturing or production and after-sales services. We present a data dimensionality reduction model for product data during different phases of the manufacturing cycle.

2 - An Optimal Algorithm for Solving a Ring-Mesh Topology Design Problem of the Optical Networks Youngjin Kim, Korea University, Anam Dong Sungbuk Ku, Seoul, 136-701, South Korea, [email protected], Jeonghee Han, Youngho Lee, Youngwook Kim, Kyunam Chang, Hyunjung Yun
In this paper, we develop an optimal algorithm to solve a ring-mesh topology design problem arising from the deployment of emerging optical networks. We investigate the dynamic cut generation procedure that combines the benefit of the column generation procedure with the efficiency of Lagrangian dual approach. Computational results of the lower and upper bounding procedures demonstrate the effectiveness of our solution procedure.

3 - Customer Targeting Analysis with Support Vector Machines and Neural Networks: A Comparative Study Yaquan Xu, Assistant Professor, Virginia State University, Department of CIS, 1 Hayden Drive, Petersburg, VA, 23806, United States, [email protected], Haibo Wang
One of the key problems in customer targeting is to find a better way to identify and profile of households who are most likely to be interested in a particular product or service. Recent studies have shown that AI methods achieved better performance than traditional statistical methods. This article introduces SVM machine learning technique to provide a model with better decision criteria. We used artificial neural networks guided by PCA as a benchmark to the comparison.

3 - Optimal Expansion of a Hybrid Circuit/Packet Switched Voice Telecommunications Backbone Raghavendran Sivaraman, Doctoral Student, Massachusetts Institute of Technology, Operations Research Center, 77 Massachusetts Avenue, Cambridge, MA, 02139, United States, [email protected], Lawrence Crom, Hui Liu, Rina Schneur, Roger Tobin, Donald Smith
Verizon transports voice telephone calls on a mixed circuit-switched and packetswitched network. We describe a routing optimization tool that minimizes the capital expended annually on network growth. In the first year, expansion according to the optimizer will cost about $5M, as compared to $9M using a predecessor tool that does not optimize. Over a five-year period, capital expenditures associated with the optimizer’s expansion plan should cost 30% less than those of the predecessor tool.

4 - On the Performance of K-Means Type Clustering Algorithms Altannar Chinchuluun, University of Florida, Weil 303, ISE, University of Florida, Gainesville, FL, 32611, United States, [email protected], Panos Pardalos
Clustering analysis plays an important role in scientific research and commercial applications. In the popular K-means algorithm, the locally optimal solution is sensitive to the initial points choice. Hence, we present two methods to find refined starting points. We compare these two methods with other refining starting condition methods through the applications of these methods to the Kmeans clustering algorithm based on their performances for various data sets.

4 - Optimal Location of Session Border Controllers in VoIP network Peter Kubat, Distinguished Member of Technical Staff, Verizon Laboratories, 40 Sylvan Road, Waltham, MA, 02451, United States, [email protected], Michael Weintraub
Session Border Controller (SBC) is a device which enables carriers to deliver feature-rich VoIP services securely and effectively across IP transport networks. In this talk we describe traffic modeling efforts and practical optimization technique used to select optimal location of SBCs in a metropolitan network.

5 - A Heuristic Aggregation Algorithm to Generate Geographical Business Clusters Bin Zhang, IBM China Research Laboratory, Building 19 Zhongguancun Software Park, Beijing, China, [email protected], Minghua Zhu, Wenjun Yin, Jin Dong
Spatial clustering technologies are widely used in market analysis. However, most proposed algorithms just considered the spatial positions and the presences of obstacles and facilities. In this paper, a heuristic aggregation algorithm is presented to model the key economic and demographic factors to generate geographical business clusters for banking market analysis. The experiments show that our approach can discover the natural characters of business distribution.

102

INFORMS PITTSBURGH— 2006 I SC52
Topics in Revenue Management and Pricing
Sponsor: Revenue Management & Pricing Sponsored Session
Chair: Alper Sen, Assistant Professor, Bilkent University, Department of Industrial Engineering, Bilkent, Ankara, 06800, Turkey, [email protected] 1 - Dynamic Pricing and Inventory Management under Advance Selling Michelle M.H. Seref, Doctoral Student, Operations Management, Decision Information Sciences, University of Florida, Stuzin Hall 361-E, Gainesville, FL, 32611, United States, [email protected], Aydin Alptekinoglu, Selcuk Erenguc
Advance selling allows customers to purchase a product prior to consumption. Customers decide whether to advance purchase or wait until the spot period based on their expected future valuation, the advance sales price and the spot price. We consider a single inventory order decision made at a deterministic leadtime during the advance sales period. We seek to find the optimal order quantity, dynamic advance sales pricing policy, and spot price that maximize the firm’s total expected profit.

SC54

I SC53
Pricing and Revenue Management: New Models
Sponsor: Revenue Management & Pricing Sponsored Session
Chair: Itir Karaesmen, Assistant Professor of Management Science, University of Maryland, Robert H. Smith School of Business, Decision and Information Technologies, College Park, MD, 20742, United States, [email protected] 1 - Profit Smoothing and Product Pricing Mahesh Kumar, Assistant Professor, Rutgers Business School, 180 University Avenue, Newark, NJ, 07102, United States, [email protected], Suresh Govindaraj
We propose a new explanation for price discrimination in products and services that is based on the sellers’ objective to smooth profits. We show that when sellers carry multiple products, the spread in buying and selling prices of any product depends on the profits and inventory positions of the other products. Consequently, sellers may price the same product differently depending on the range of products that they carry.

2 - Dynamic Pricing of Perishables with Renewal Demands and Menu Costs Emre Berk, Bilkent University, Faculty of Business Administration, Bilkent University, Ankara, 06800, Turkey, [email protected], Ulku Gurler
We consider dynamic pricing of perishable assets in the presence of pricesensitive renewal demand processes and non-negligible price change costs with the objective of maximizing the disounted expected profit for an initial inventory of Q items. We establish some structural results on the behavior of the objective function with respect to the selling price and the horizon length. We also present some numerical results.

2 - Centralized vs. Decentralized Supply Chains Ann Chan, Assistant Professor, Virginia Tech, 250 Durham Hall, Blacksburg, VA, 24061-0118, United States, [email protected]
When customer demand is stochastic and price sensitive, we study the loss of expected profit due to double marginization. We also analyze and compare the supply chain performances under different operations mechanisms: buy-back, inventory sharing, coordination and competition.

3 - Dynamic Pricing with Strategic Customers: The Role of Customer Heterogeneity Xuanming Su, Assistant Professor, University of California, Haas School of Business, Berkeley, CA, 94720, United States, [email protected]
We consider a firm who prices and rations a finite inventory of products over a finite time horizon, and consumers who strategically time their purchases under rational expectations of future prices and availability. The optimal policy depends critically on customer heterogeneity along two dimensions: their willingness to pay, and their willingness to wait.

3 - Stochastic Models for Joint Pricing and Revenue Management Ayse Kocabiyikoglu, Assistant Professor, Bilkent University, Department of Business Administration, Bilkent, Ankara, 06800, Turkey, [email protected], Ioana Popescu
We extend the traditional revenue management paradigm to simultaneously optimize price and capacity allocation decisions, under a stochastic price dependent demand model. We provide conditions for the monotonicity of the capacity allocation policy, as well as the uniqueness of the optimal solution. We provide bounds on the optimal revenue and the results of numerical experiments, which demonstrate significant improvements in revenue from joint price-capacity allocation optimization.

4 - Market Level Price Optimization for Hotels and Gaming Resorts Ahmet Kuyumcu, [email protected], Amar Duggasani
Many hotel chains and gaming resorts own multiple properties in close proximity and need abilities to shift demand to optimize their profits. This requires new control capabilities such as free upgrades, cross-sells, and down-sells. In addition, forecasting, overbooking, and optimization models could utilize market-level information. This presentation discusses challenges and opportunities in implementing a market-level price optimization approach.

4 - Optimal Bundle Formation and Pricing of Two Products with Limited Stock Alper Sen, Assistant Professor, Bilkent University, Department of Industrial Engineering, Bilkent, Ankara, 06800, Turkey, [email protected], Ulku Gurler, Salih Oztop
We consider a retailer that sells a fixed stock of two products not only as independent items but also as a bundle over a season. The retailer decides how many bundles to form and what prices to charge for the bundle so that expected profits over season are maximized. Product demands follow a Poisson Process and customer reservation prices are assumed to have a joint distribution. We study the impact of reservation prices, inventory levels, arrival rates and bundle formation costs.

I SC54
Panel Discussion: Integrated Operations and Marketing for Servng Emerging Markets
Cluster: Operations and Marketing for Emerging Markets Invited Session
Chair: Vish Krishnan, Professor, University of California - San Diego, Rady School of Management, 9500 Gilman Drive MC 0093, La Jolla, CA, 92093, United States, [email protected] Paul Griffin,Professor, Georgia Institute of Technology, 765 Ferst Drive, Atlanta GA 30332, United States, [email protected], Christina Scherrer, Julie Swann 1 - Panel Discussion: Integrated Operations and Marketing for Serving Emerging Markets Moderator: Vish Krishnan,Professor, University of California - San Diego, Rady School of Management, 9500 Gilman Drive MC 0093, La Jolla CA 92093, United States, [email protected], Panelist: Cheryl Druehl, Uday Karmarkar, Sunder Kekre, Kamalini Ramdas, Aleda Roth
Emerging markets present new opportunities and challenges for global firms. Their size and limited penetration present growth potential. However, infrastructural and budget constraints could challenge margins and even realization of growth. This panel seeks to discuss how integrating the operations and marketing decisions can offer firms the opportunity to realize the potential of emerging markets.

103

SC55 I SC55
Supply Chain Management I
Contributed Session

INFORMS PITTSBURGH — 2006

Sunday, 4:30pm - 6:00pm
I SD01
Optimization on Graphs
Sponsor: Optimization/ Network and Combinatorial Optimization Sponsored Session
Chair: Benjamin Van Roy, Associate Professor of Management Science and Engineering and of Electrical Engineering, Stanford University, Terman Engineering Center, Stanford, CA, 94305, United States, [email protected] 1 - Message-Passing Algorithms for Optimization Ciamac Moallemi, Graduate Student, Stanford University, Department of Electrical Engineering, Packard 274, Stanford, CA, 94305, United States, [email protected], Benjamin Van Roy
A nonserial dynamic program is an optimization problem with an objective that decomposes according to a graphical structure. Over the past few years, there has been a growing interest in message-passing algorithms for the solution of such problems, however these methods are still poorly understood and have received little attention in the OR community. We will present such an algorithm in several settings and discuss convergence results.

Chair: A. M. M. Jamal, Professor, Department of Management, Box 10350, Southeastern Louisiana University, Hammond, LA, 70402, United States, [email protected] 1 - Coordinating Product Development with Suppliers and Performance Impacts Yunsook Hong, PhD Candidate, Arizona State University, Supply Chain Management Department, W. P. Carey School of Business, Tempe, AZ, 85287-4706, United States, [email protected], John Pearson
As manufacturers turn to their suppliers for designing components, they face the problem of coordinating their inter-organizational product development efforts. We identify three strategies for manufacturers to coordinate component developments with multiple suppliers. Using a survey research methodology, we investigate the impacts of the coordination strategies on product development performance.

2 - SCM and NPD Fouad El Ouardighi, Associate Professor, Essec Business School, BP 105, Cergy Pontoise, 95021, France, [email protected], Bowon Kim
In this paper, we develop a dynamic model of collaboration between a manufacturer and its supplier, where the fundamental issue is, for each player, how to allocate own resources between improving an existing product and developing a new one. We study the optimal time path of effort allocation for each player in the non-cooperative setting, and then look into the cooperative setting.

2 - Lower Bounds and Metaheuristics for a Bounded Vertex Coloring Problem Enrico Malaguti, DEIS - University of Bologna, viale Risorgimento, 2, Bologna, 40136, Italy, [email protected], Albert Einstein Fernandes Muritiba, Manuel Iori, Paolo Toth
We consider a bounded version of the Vertex Coloring Problem, where each vertex i has an associated positive weight w(i), and the total weight of the vertices assigned to the same color is bounded. The problem can be also seen as a Bin Packing Problem subject to compatibility restrictions. We propose new lower bounding procedures and an effective metaheuristic approach for the problem.

3 - An Integrated Framework Matching Product Architecture with Supply Chain Design Policies Femi Famuyiwa, Rumble Fellowship, Wayne State University, 4815 4th Street, Industrial and Manufacturing Engineering, Detroit, MI, 48201, United States, [email protected], Leslie Monplaisir
While it is widely believed that it is better to consider supply chain concerns during product development stage, however, there are few existing studies that provide a quantitative framework for making such decisions early in the architecture selection stage. This paper tries to fill this gap by proposing a quantitative framework. The optimization model will incorporate decision variables so that one can examine the impact of modularity decisions on supply chain policies.

3 - (Fraction) Dimension of a Poset and Constraint Feedback Arc Set Problem Kamal Jain, Senior Researcher, Microsoft Research, One Microsoft Way, Redmond, WA, 98052, United States, [email protected], Rajneesh Hegde
The dimension of a partially ordered set (poset) is the minimum integer k such that the partial order can be expressed as the intersection of k total orders. We prove it is hard to approximate the dimension of a poset with n elements within a factor n {0.5 - \epsilon} for any \epsilon ccc 0. The same hardness of approximation holds for the fractional version of poset dimension. On a positive side we consider a minimum “constrained” feedback arc set problem in tournament.

4 - A Hybrid Solution Procedure for Design for Supply Chain (DFSC) Problems Nuri Gokhan, PhD Student, University of Pittsburgh, Industrial Engineering, 1048 Benedum Hall, Pittsburgh, PA, 15261, United States, [email protected], Kim Needy, Bryan A. Norman
This research investigates integrating product and supply chain (SC) design to maximize benefits throughout the product life cycle. Product design and SC performance are modeled in a multi-objective fashion. Differences between sequentially and simultaneously optimizing the product and SC design are analyzed. A hybrid solution procedure including MIP, GA, Tabu Search, and simulation is proposed. Preliminary models and results will be presented including solution quality and computational time.

I SD02
Joint Session Open-Source/ICS: New in COIN-OR
Cluster: Open-Source Software: Open Source, Open Standards, Open Data Invited Session
Chair: Brady Hunsaker, Assistant Professor, University of Pittsburgh, Department of Industrial Engineering, 1036 Benedum Hall, Pittsburgh, PA, 15261, United States, [email protected] 1 - Cut Generator for Chvatal-Gomory Rank 1 Inequalities Brady Hunsaker, Assistant Professor, University of Pittsburgh, Department of Industrial Engineering, 1036 Benedum Hall, Pittsburgh, PA, 15261, United States, [email protected], Mustafa Baz, Craig Tovey
We present code submitted to the COIN-OR Cut Generator Library (CGL) for Chvatal Gomory rank 1 inequalities. Separation of these inequalities requires the solution of a MIP and has been explored by several researchers. We discuss the formulation used to separate the inequalities, user interface and parameters, and possible applications. We also discuss the potential use of this code as a model for other MIP-based CGL routines.

5 - Modeling Daily Exchange Rate Movements A. M. M. Jamal, Professor, Department of Management, Box 10350, Southeastern Louisiana University, Hammond, LA, 70402, United States, [email protected], Shakil Quayes
The value of international currencies traded daily is approximately 2 trillion dollars (Bank of International Settlements). In comparison, the value of goods and services would be less than 5 percent of that amount. Most econometric models of exchange rates are based on monthly or quarterly data. We will develop an econometric model for exchange rates as affected by the daily fluctuation in the factors such as interest rates and the volume of financial transactions.

2 - LEMON: A Library of Efficient Models and Optimization in Networks Alpár Jüttner, Eötvös University, Budapest, Pázmány Péter sétány 1/C, Budapest, H-1117, Hungary, [email protected]
LEMON is an open source library written in C++ language. It provides a set of efficient and easy-to-use implementation of common data structures and algorithms in the area of optimization and helps implementing new ones. The main focus is on graph structures and graph related optimization algorithms, such as shortest paths, network flows, connectivity and matching algorithms etc. LEMON also provides a simple but powerful file format to store graphs along with attributes assigned to its elements.

104

INFORMS PITTSBURGH— 2006
3 - Components for Applied OR: k-Shortest Path Mary Helander, IBM Research, TJ Watson Research Center, 1101 Kitchawan Road, Yorktown Heights, NY, 10598, United States, [email protected]
The goal of this COIN-OR project is to share codes for published k shortest path algorithms, which are often building blocks of practical OR applications in transportation, vehicle routing, facility location, telecommunication network design, social network anlaysis, and other areas. The seed contribution includes a C code for the Miaou and Chin EJOR 1991 k shortest path algorithm.

SD05

1 - The Martingale Approach to Operational and Financial Hedging Rene Caldentey, Assistant Professor in Operations Management, Stern School of Business, New York University, 44 West Fourth Street, KMC 8-77, New York, NY, 10012, United States, [email protected], Martin Haugh
We describe an approach for the optimal selection of dynamic operating and financial hedging strategies when the decision maker is risk averse. Risk aversion is imposed through constraints on the feasible policies. We apply the approach to some standard operations problems.

I SD03
Variational Inequalities, MPECs and Maximum Likelihood Estimation
Sponsor: Optimization/ Linear Programming and Complementarity Sponsored Session
Chair: Daniel Ralph, Reader in Management Science, University of Cambridge, Judge Business School, Cambridge, CB2 1AG, United Kingdom, [email protected] 1 - Solving Asymmetric Variational Inequalities via Convex Optimization Michele Aghassi, [email protected], Georgia Perakis, Dimitris Bertsimas
Using duality, we reformulate the asymmetric variational inequality (VI) problem over a conic region as a single-level, and many-times continuously differentiable optimization problem. We give sufficient conditions for the convexity of this reformulation and identify a class of VIs, of which monotone affine asymmetric VIs over polyhedra are a special case, and can be solved using widely-available and commercial-grade convex optimization software.

2 - Interplay Between Operational Flexibility and Financial Hedging Jan Van Mieghem, Harold L. Stuart Distinguished Professor of Managerial Economics, Professor of Operations Management, Kellogg School of Management, Northwestern University, MEDS Department, 2001 Sheridan Road, Evanston, IL, 60208-2009, United States, [email protected], Jiri Chod, Nils Rudi
We investigate the relationship between operational flexibility and financial hedging. Flexibility influences (i) the firm’s need to hedge by affecting profit variance, (ii) its ability to hedge by affecting hedging efficacy, and (iii) its valuation of hedging by increasing the firm’s wealth. We show that flexibility and hedging tend to be complementary (substitutes) in reinforcing (mitigating) each other’s marginal effect on the mean-variance criterion (expected exponential utility).

3 - Capacity Investment in Imperfect Markets: The Interaction of Operational and Financial Decisions Onur Boyabatli, Doctoral Candidate, INSEAD, Boulevard de Constance, Fontainebleau, 77305, France, [email protected], Beril Toktay
This paper analyzes the interaction between the operational (technology choice, capacity investment, production quantity) and the financial (financial risk management and financing) decisions. We model the interplay between a firm that can raise external capital and a creditor that determines the cost of the external capital within imperfect capital markets. We demonstrate that some of the common operational insights do no continue to hold in imperfect markets.

2 - On a Multiclass Spatial Equilibrium Problem Patrice Marcotte, Professor, University of Montreal, CP 6128, Succ. Centre-Ville, Montreal, QC, H3C3J7, Canada, [email protected], Michel Gendreau
We consider a multiclass extension of the spatial price equilibrium model, where each user class is characterized by its “value-of-time”. We formulate the equilibrium conditions as an asymmetric and non-monotone variational inequality, for which we propose a decomposition framework that efficiently exploits the problem’s structure, as is demonstrated on large-scale instances. Finally, we discuss the MPEC that consists in setting the prices of a leader firm in this context.

I SD05
Probabilistic and Dominance-Constrained Optimization
Sponsor: Optimization/ Stochastic Programming Sponsored Session
Chair: Andrzej Ruszczynski, Professor, Department of Management Science and Information Systems, Rutgers University, 94 Rockefeller Road, Piscataway, NJ, 00854, United States, [email protected] 1 - Bounds on the Values of Derivatives Under Partial Knowledge of the Asset Price Distribution András Prékopa, Professor, RUTCOR, Rutgers University, 640 Bartholomew Road, Piscataway, NJ, 08854, United States, [email protected]
Bounding the value of a financial derivative comes up either because the asset price distribution is only partially known or to compute the value is hard. New bounding methods will be presented that can be used both in the single and multiple asset cases in connection with European and American options. Moment information or knowledge of the univariate marginals of distributions of asset prices are assumed to be available.

3 - A New Optimization Approach to Maximum Likelihood Estimation of Structural Models Che-Lin Su, Post-Doctoral Research Fellow, CMS-EMS, Kellogg, Northwestern University, 580 Leverone Hall, 2001 Sheridan Road, Evanston, IL, 60208, United States, [email protected], Kenneth Judd
Maximum likelihood estimation of structural models is regarded as computationally difficult. This impression is due to a focus on the Rust’s Nested Fixed-Point (NFXP) approach. We present a direct optimization approach to the problem and show that it is significantly faster than the NFXP approach when applied to the canonical Zurcher bus repair model. The direct optimization approach is also applicable to other structural estimation problems such as Nash games, auctions and RBC models.

4 - Some Developments in Stochastic MPECs Daniel Ralph, Reader in Management Science, University of Cambridge, Judge Business School, Cambridge, CB2 1AG, United Kingdom, [email protected], Huifu Xu
We study stochastic mathematical programs with equilibrium constraints, SMPECs, where the second stage is a complementarity problem that is parametric in the first stage variables. Applications appear in energy markets. Our aim is to understand how the regularisation technique in MPECS, studied by Scholtes and others, might be adapted to sample average approximation SMPECs. Our convergence the analysis is a departure in that it makes use of MPEC multipliers. (With Huifu Xu, Southampton, UK)

2 - Optimization Subject to Stochastic Ordering Constraints Darinka Dentcheva, Professor, Stevens Institute of Technology, Mathematical Sciences, Castle Point on Hudson, Hoboken, NJ, 07030, United States, [email protected], Andrzej Ruszczynski, Rene Henrion
We discuss stochastic optimization problems involving a continuum of probabilistic (chance) constraints. These are called also stochastic dominance of first order or stochastic ordering constraints. Necessary and sufficient conditions of optimality will be reviewed. We focus on the stability of the problem when the underlaying probability distributions are subject to perturbations, e.g., when the distributions are estimated. At the end we sketch several applications of the model.

I SD04
OM/Finance Interface
Cluster: Operations and Marketing for Emerging Markets Invited Session
Chair: Beril Toktay, Associate Professor, College of Management, Georgia Institute of Technology, 800 West Peachtree Street, NW, Atlanta, GA, 30308, United States, [email protected]

105

SD06

INFORMS PITTSBURGH — 2006
2 - A Comprehensive Computational Study of Scaling Techniques for Linear Programming Joe Elble, Department of Computer Science, University of Illinois Urbana-Champaign, Urbana, IL, United States, [email protected], Nick Sahinidis
The scaling of linear programs, while poorly understood, is not devoid of techniques. Scaling is the most commonly used preconditioning technique in linear programming solvers. This paper presents a review and computational comparison of techniques for obtaining scaling factors for linear systems. Old and very recent scaling techniques are investigated.

3 - Methods for Solving Optimization Problems with Stochastic Dominance Constraints Andrzej Ruszczynski, Professor, Department of Management Science and Information Systems, Rutgers University, 94 Rockefeller Road, Piscataway, NJ, 00854, United States, [email protected], Nilay Noyan, Gabor Rudolf
For stochastic optimization problems involving dominance constraints of the second order, using the theory of existence of measures with given marginals, we derive specialized column generation techniques. Next, for problems with first order constraints, we develop specialized lower and upper bounds, based on conditional second order constraints, and we construct near-optimal feasible solutions. In portfolio optimization the methods yield improvements of order of magntide.

I SD06
Algorithmic Topics in Constraint Programming
Sponsor: INFORMS Computing Society/Constraint Programming Sponsored Session
Chair: Irit Katriel, Professor, BRICS, University of Aarhus, Aabogade 34, Aarhus, AA, Denmark, [email protected] 1 - The CP(Graph) Computation Domain in Constraint Programming Grégoire Dooms, Université Catholique de Louvain, INGI, place St Barbe, 2, Louvain-la-Neuve, 1348, Belgium, [email protected], Yves Deville, Pierre Dupont
CP(Graph) introduces graph variables which domain is abstracted as a graph interval and constraints over these variables. It is integrated with the finite domain and finite sets domain, allowing the reuse of constraints of these domains. Graph intervals ease the formulation and implementation of filtering algorithms for global graph constraints. We present novel graph constraints and a proof of concept available as contributed open-source code in the Gecode constraint programming library.

3 - Effective Multistart for Reaching Feasibility in Difficult Nonlinear Programs John Chinneck, Professor, Carleton University, Systems & Computer Engineering, Ottawa, ON, K1S 5B6, Canada, [email protected], Matthew MacLeod
Finding a feasible solution for a set of nonlinear constraints can be very difficult. Local solvers may fail if not started close to a feasible region, so an expensive multistart approach may be needed. Instead of using the costly local solver each time, we use cheap Constraint Consensus methods to explore the variable space. We apply a local solver only when confident that we are close to a feasible region. Experimental results are very promising.

I SD08
Sponsor: Technology Management Sponsored Session
Chair: David Moore, Director of klicnet.org, President of Knowledge Logic LLC, 3788 Davidson Place, Boulder, CO, 80305, United States, [email protected] 1 - Pharmaceutical-Biotechnology R&D: Implications for Alliance Performance Jongwook Kim, Assistant Professor, Western Washington University, 351 Parks Hall MS9075, 516 High Street, Bellingham, WA, 98225, United States, [email protected]
Alliance opportunities are often characterized by information asymmetry, particularly where key resources are intellectual property. In the context of biotechnology-pharmaceutical alliances, I test how alliance characteristics impact performance. The data support the claim that utilization of network ties may be informational in nature where the role of reputation and network ties in latter stages diminish as the odds of success are higher, and more emphasis is placed on firm competence.

2 - Grammar Constraints Meinolf Sellmann, Assistant Professor, Brown University, 115 Waterman Street, PO Box 1910, Providence, RI, 02912, United States, [email protected]
We introduce constraints based on grammars in the Chomsky hierarchy. We devise an arc-consistency algorithm for context-free grammars, investigate when logic combinations of grammar constraints are tractable, show how to exploit non-constant size grammars and re-orderings of languages, and study where the boundaries run between regular, context-free, and context-sensitive grammar filtering.

3 - Set Solvers Revisited Carmen Gervet, Brown University, Box 1910, Providence, RI, 02912, United States, [email protected], Pascal Van Hentenryck
The language of sets naturally models combinatorial designs. Our findings show that by designing set solvers in analogy with Finite Domain solvers, we can strengthen propagation effectively. A set domain is now totally ordered by length-lex ordering that directly encodes cardinality and lexicographic information essential in set problems. The solver enforces bound-consistency on all unary constraints in time O(k log k), where k is the set cardinality. The approach can be applied to multisets.

2 - Managing Pre-Technological Knowledge: A Multi-Dimensional Approach Charles Weber, Assistant Professor of Engineering and Technology Management, Portland State University, ETM, PO Box 751, Portland, OR, 97207, United States, [email protected]
Bohn (1994) states that many high technology industrial processes are based on pre-technological knowledge — knowledge that is incompletely characterized or cannot be measured. An empirical study suggests that high tech industrial processes are based on on subsystem knowledge and prior, architectural knowledge that has been completely characterized. However, integration knowledge is pre-technological, and may be inherently so.

I SD07
Computational Approaches for Large Scale Optimization Problems
Sponsor: INFORMS Computing Society/Optimization: Computational Optimization and Software Sponsored Session
Chair: J. Cole Smith, Associate Professor, University of Florida, ISE Department, 303 Weil Hall, Gainesville, FL, 32611-6595, United States, [email protected] Co-Chair: Gino Lim, Assistant Professor, University of Houston, 4800 Calhoun Road, E211, Engineering Building 2, Houston, TX, 77204, United States, [email protected] 1 - Solving Large Scale Stochastic Programs Using Resampling in Stochastic Decomposition Suvrajeet Sen, Professor, Department of Systems and Industrial Engineering, University of Arizona, Tucson, AZ, 85721, United States, [email protected], Zhihong Zhou, Kai Huang
Stochastic Decomposition provides one of the most effective methods for solving two-stage stochastic programs. We propose an extension that allows cuts based on re-sampling from a subset of outcomes that have been revealed in the past. Convergence results, and computational experiments will be reported with this new version of Stochastic Decomposition.

3 - Patents in Practice: Systemic Failures of Knowledge Representation in the US Patent System Tony Briggs, Doctoral Candidate, Boston University Graduate School of Management, Information Systems Department, 595 Commonwealth Avenue, Boston, MA, 02215, United States, [email protected], Paul Carlile
The US patent system is charged to promote the progress of science and useful arts. While patents are managed as economic devices across firms, they are seldom used as knowledge repositories to shape innovation within firms. We use a knowledge management framework to examine 3 generic patenting practices: development, examination, and enforcement. We find that different practices lead to either knowledge obfuscation or obsolescence, resulting in the systemic failure to inform future innovation.

4 - Social Capital and the Creation of Knowledge Claudia Gonzalez-Brambila, Business School, Instituto Technologico Autonomo, Rio Hondo 1, Mexico, D.F., 01000, Mexico, [email protected], Krackhardt David, Francisco Veloso
We examine the relation between knowledge creation, measured through published papers, and social capital, characterized through co-authorship. Using an extensive panel, analysis suggests that, contrary to previous results, structural holes are not significant; what matters are direct ties, being central, working across areas of knowledge, and being in non dense networks.

106

INFORMS PITTSBURGH— 2006
5 - Organizational Learning in Distributed Innovation Planning Edward Anderson, Professor, University of Texas, McCombs Business School, 1 University Station B6500, Austin, TX, 78733, United States, [email protected], Nitin Joglekar
Using a complex systems/system dynamics perspective, we examine the coevolutionary relationship between the market, innovation, products and capabilities in distributed environments. In particular, we argue that the role of capability planning is of the highest leverage in guiding innovation. We also argue that this planning must incorporate extensive levels of risk management and flexibility because of the inherent path dependence in such complex systems.

SD11

2 - Minitab, Inc. - Response Surface Optimization Using Minitab Nicole DeFazio, Minitab, Inc., 1829 Pine Hall Road, State College, PA, 16801, Eduardo Santiago
Attendees will learn how to use Minitab Statistical Software to perform Response Surface Optimization. Tips and tricks, customizing Minitab for teaching and business as well as additional advanced capabilities will be demonstrated. Minitab Statistical Software is the ideal package for Six Sigma and other quality improvement projects and is used in over 4000 colleges and universities worldwide for the past 30 years.

I SD11
BUGS: Using OR to Study Baseball, Umpiring, Golf and Soccer
Sponsor: OR in Sports Sponsored Session
Chair: Bruce Bukiet, Associate Professor of Math Sciences, New Jersey Institute of Technology, Department of Math Sciences, Newark, NJ, 07102, United States, [email protected] 1 - A Markov Model of Baseball: Applications to Two Sluggers Mark Pankin, President, MDP Associates LLC, 1018 N. Cleveland Street, Arlington, VA, 22201, United States, [email protected]
The progress of an inning of baseball is a natural process for a Markov model. The author will discuss one he has developed that contains quite a few detailed facets of the game. One application is about the effects of the St. Louis Cardinals batting the pitcher eighth in the second half of the 1998 season in an attempt to help Mark McGwire set the new single season home run record. The second applies the model to determine when the Giant’s Barry Bonds in 2001-02 should be walked intentionally.

I SD09
Sciences of the Artificial Since Simon 1969: Design, Problem Solving, and System Improvement in Management
Cluster: New Product Development Invited Session
Chair: Karl Ulrich, Professor, The Wharton School, 500 Huntsman Hall, Philadelphia, PA, 19104, United States, [email protected] 1 - Product Development as a Problem Solving Process Christian Terwiesch, Associate Professor, University of Pennsylvania, 548 JMHH, Philadelphia, PA, 19104, United States, [email protected]
We establish a problem solving framework that towards product development. Our problem solving view, which goes back to the work by Simon, is sufficiently general that it allows us to abstain from favoring any one of the many existing academic disciplines associated with product development. It also reveals many similarities between the existing literature streams, illustrates how these streams have cross-fertilized each other, and points to fruitful areas of future research.

2 - Knowledge, Causality, and Hierarchy in System Design Roger Bohn, Professor, University of California - San Diego, 9500 Gilman Drive #0519, IR/PS, La Jolla, CA, 92093-0519, United States, [email protected]
Although we design using hierarchies and modules, products and systems don’t actually behave accordingly. Knowledge about them can be described using causal graphs, and these graphs generally contain extensive linkages among nominally independent subsystems. Such knowledge is hard to learn, but critical to good design.

2 - The Traveling Umpire Problem Hakan Yildiz, Doctoral Student, Carnegie Mellon University, Tepper School of Business, Pittsburgh, 15213, United States, [email protected], Michael Trick
We introduce The Traveling Umpire Problem (TUP) as a multi-objective sports scheduling problem. Like the Traveling Tournament Problem for league scheduling, the TUP is based on the most important features in scheduling the umpires for Major League Baseball. We explore integer and constraint programming approaches to this problem, and show a relationship to graph coloring.

3 - Creating New Enterprises as a Structured Design Problem Karl Ulrich, Professor, The Wharton School, 500 Huntsman Hall, Philadelphia, PA, 19104, United States, [email protected]
Enterprises can be represented by a template or grammar using standard primitives (e.g., products, customers, transactions). By representing enterprises this way, a search space of new ventures can be made explicit and structured design methods can be used to explore this space. This framework, with its roots in Simon’s Sciences of the Artificial offers the potential for a rigorous theoretical foundation to the enterprise creation process in the fields of entrepreneurship and innovation.

3 - Golf Handicap: A Moving Average of Moving Order Statistics Tom Spencer, Professor, Walden University, School of Management, United States, [email protected], Ivan Zorych
A golf handicap can be viewed as a moving average of moving order statistics. A brief review of existing results based on the naïve assumption of i.i.d. normality of the generating sequence (i.e. your golf scores) will be given, along with a discussion of our simulations using a large multi-year data base of handicapped scores from an industrial Golf Tournament League.

4 - Modeling Tactical Changes of Formation as a Non-Zero-Sum Game in Soccer Nobuyoshi Hirotsu, Lecturer, Juntendo University, 1-1 Hiragagakuendai, Inba, Chiba-ken, 270-1695, Japan, [email protected], Chikara Miyaji
In general, a sport game played by two teams or two players can be considered as a zero-sum game for the purpose of gaining a win. However, the normal system used by soccer leagues is to award three points for a win, one point for a draw and none for a defeat. From the standpoint of gaining league points, a soccer game can be modeled as a non-zero-sum game. Here, we propose a game theoretic approach to modeling tactical changes of formation. We show a solution of the game with the expected number of league points and discuss a merit of cooperation between two teams, using real data of the J League.

I SD10
Software Demonstration
Cluster: Software Demonstration Invited Session
1 - JMP - The Joy of Data Analysis with JMP Curt Hinrichs, SAS Institute, SAS Campus Drive, Cary, NC, 27513, [email protected]
JMP is a state of the art statistical package designed for the busy professional who wants to quickly and easily analyze their data and who needs sophisticated analyses - from classical statistical methods to modern design and exploratory data mining - that only SAS can provide. Intuitive, interactive and graphical, JMP lets you focus on the insight your data can provide. We will cover some of the powerful features and intuitive navigation of JMP.

107

SD12 I SD12

INFORMS PITTSBURGH — 2006
sections taught completely on-campus in a traditional format. All were administered equivalent examinations. Student feedback was mixed. Suggestions and strategies for the future are discussed.

Nicholson Student Paper Prize Competition, II
Cluster: Nicholson Student Paper Prize Invited Session
Chair: Chung-Piaw Teo, Associate Professor, National University of Singapore, Business School, 21 Lower Kent Ridge Road,Singapore 119077, Singapore, [email protected] 1 - Coordinating Efforts of Multiple Retailers in a Decentralized Supply Chain Stephen Shum, PhD Candidate, Massachusetts Institute of Technology, 77 Massachusetts Avenue, E40-130, Cambridge, MA, 02139, United States, [email protected], David Simchi-Levi
We study the impact of effort in a supply chain with multiple retailers. The costly effort engaged by a retailer may increase or decrease the demands of other retailers. We show that traditional contracts coordinate only under very restrictive conditions. We propose revenue sharing with fixed target rebate. These contracts are shown to be both coordinating and flexible. Interestingly, similar contracts have been used in the fresh food industry to motivate retailers to promote the products.

3 - Integrating Systems Engineers into an Operations Research Capstone James Lowe, Professor, 2354 Fairchild Drive, USAF Academy, CO, 80840, [email protected]
In 2006, USAFA graduated its first class of Systems Engineering majors. The students of this major were placed into several different capstone efforts. SE majors were added to the long-standing OR capstone effort with positive, yet mixed, results. We address the advantages, challenges, and future plans.

4 - Scenario-Based Examinations in a Core OR Course Andrew Armacost, Associate Professor of OR, US Air Force Academy, 2354 Fairchild Drive, Suite 6J100, USAFA, CO, 80840, United States, [email protected]
In our core course exams, we have found that students often spend a large amount of time on trying to understand the written scenarios (which is imperative), and less time applying modeling techniques and interpreting solutions. We describe two approaches we’ve used to allow all three elements to occur. Both approaches involve providing scenarios in advance in a manner that encourages students to understand the scenarios before the exam and to anticipate questions that might occur on the exam.

2 - Supply Chain Coordination and Influenza Vaccination (Submitted for the 2006 Pierskalla Award) Hamed Mamani, PhD Student, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Building E40-130, Cambridge, MA, 02139, United States, [email protected], David Simchi-Levi, Stephen Chick
Billions of dollars are being allocated for influenza pandemic preparedness. Vaccination is a primary weapon for fighting influenza outbreaks. Differences that distinguish the influenza vaccine supply chain from the typical supply chains include a nonlinear value of sales and vaccine production yield issues. We show that production risks, taken by the manufacturer, lead to an insufficient supply of vaccine and design a variant of the cost sharing contract that can coordinate the supply chain.

I SD14
Statistical Models
Contributed Session
Chair: Connie Borror, Associate Professor, Arizona State University West, 4506 W. Kaler Circle, Glendale, AZ, 85301, United States, [email protected] 1 - Alternative to the Least Squares Regression Subhash C. Narula, Professor, Virginia Commonwealth University, School of Business, 1015 Floyd Avenue, Richmond, VA, 232844000, [email protected], John Wellington
Because the least squares regression is very sensitive to outliers, is affected by multicollinearity among the predictor (regressor) variables, and long tailed error distributions a number of alternatives procedures have been proposed. To overcome these difficulties, a number of alternatives have been proposed. Our objective is to introduce some of the available procedures that have been proposed.

3 - On the Computational Complexity of MCMC-based Estimators in Large Samples Alexandre Belloni, MIT, 77 Massachusetts Avenue, Cambridge, MA, 02139, United States, [email protected], Victor Chernozhukov
This paper studies the computational complexity of Bayesian and quasi-Bayesian estimation in large samples carried out using a basic Metropolis random walk. Our framework covers cases where the underlying likelihood or extremum criterion function is possibly non-concave, discontinuous, and of increasing dimension, but the posterior or quasi-posterior based on it approaches a normal density in large samples. Using this central limit framework to provide structural restrictions for the problem, we show that the running time of the algorithm in large samples is bounded in probability by a polynomial in the parameter dimension d, and in particular is O_p(d^3) in the leading cases. Large sample asymptotics allows to invoke the central limit theorem to bound the deviations from continuity and concavity, in a specific manner that allows to claim that the computational complexity is polynomial.

2 - Statisical Approximation of Sustainably Selective Welfare Functions Kobi Abayomi, Visiting Professor, Haverford College, PO Box 250219, Manhattan, NY, 10025, United States, [email protected]
”Sustainability”, in the economic sense, is a criteria in need of definition. Chichilnisky has proposed an axiomatic standard: that a welfare function is “sustainable” if, and only if, it is non-negligible on both the finite and infinite parts of its domain - the space of utility streams. I investigate statistical approaches to Chichilnisky’s axiomatization, which can be regarded as finite time tests of the “sustainability” of a utility, and, thusly, a development path.

I SD13
Innovations in Undergraduate Operations Research Curricula
Sponsor: Education (INFORM-ED) Sponsored Session
Chair: David Mullin, Associate Professor of Economics, U.S. Air Force Academy, HQ USAFA/DFEG, 2354 Fairchild Drive, Suite 6K-110, USAF Academy, CO, 80840, United States, [email protected] 1 - Online Approaches to Teaching Operations Research David Mullin, Associate Professor of Economics, U.S. Air Force Academy, HQ USAFA/DFEG, 2354 Fairchild Drive, Suite 6K-110, USAF Academy, CO, 80840, United States, [email protected]
This paper presents new online technology in teaching operations research with an intuitive approach in graphing standard operations research topics, such as linear programming. Graphs are automatically evaluated and graded. The system uses TeX and LoadTeX to render files that are stored in a relational database. An instructor may use pre-existing graphing problems or author new ones. This system is available through Aplia,Inc, a company that has more than 80,000 ongoing students.

3 - Joint Determination of Optimal Process Control Policy and Warranty Length in a Supply Chain Xiaowei Xu, Assistant Professor, MSIS, Rutgers Business School, Ackerson Hall, Room 208, 180 University Avenue, Newark, NJ, 07102, United States, [email protected]
We assume that the state of a production system follows a Brownian motion and the failure time of a product follows a proportional hazard model. We show that the optimal control barrier is decreasing in warranty length and the optimal warranty length is shorter in a supply chain with asymmetric information of production technology than in a supply chain with full information.

4 - Design of Experiments with Secondary Constraints Szu Hui Ng, Assistant Professor, National University of Singapore, 10 Kent Ridge Crescent, Singapore, 119260, Singapore, [email protected]
Standard experimental designs typically assume that the factor space is cubodial or spherical. However in many situations, secondary constraints may exist, causing interdependencies among the factors and resulting in an irregular operability region. The region imposed by the constraints is often unknown a priori, rendering standard experimental designs inappropriate. Here we propose an approach to design experiments in these restricted regions and study several irregularly shaped designs.

2 - Lessons Learned: On-Line Education in the Undergraduate Curriculum Christina Scherrer, Assistant Professor, Southern Polytechnic State University, 1100 S. Marietta Parkway, Atlanta, GA, 30060, United States, [email protected], Patricia Carden
We present results and insights from an experiment conducted with our department’s undergraduate introductory statistics course. In this experiment, two sections were taught completely on-line using live audio technology and two

108

INFORMS PITTSBURGH— 2006
5 - Process Capability Estimation in the Presence of Measurement Error Connie Borror, Associate Professor, Arizona State University West, 4506 W. Kaler Circle, Glendale, AZ, 85301, United States, [email protected]
Confidence intervals are constructed on two measures of process capability, Cp and the dpm, to demonstrate the importance of considering variability present due to the measurement error. The process requires a gauge R&R study be performed prior to the collection of data for the estimation of process capability. The procedure and results will be illustrated using an example involving a twofactor model commonly encountered in gauge R&R studies.

SD17

The bidding dynamics during an auction play an important role in explaining auction outcome, yet remain underinvestigated due to a lack of common process metrics and analytical methods. We propose several such metrics that enable a comparison of bidding dynamics across auctions. We also show that techniques from Functional Data Analysis (FDA) yield additional insights into bidding dynamics. We illustrate our arguments with an analysis of 700+ reverse auctions in the automotive industry.

I SD16
Advances in Revenue Management
Cluster: Dynamic Pricing and Forecasting Invited Session
Chair: Guillaume Roels, Assistant Professor, UCLA Anderson School of Management, 110 Westwood Plaza, Los Angeles, CA, 90077, United States, [email protected] 1 - Hard Block Code-Sharing Contracts for Airline Revenue Management Guillermo Gallego, Professor, Industrial Engineering and Operations Research, Columbia University, New York, NY, United States, [email protected], Youyi Feng
We consider capacity contracts between a marketing and operating airline where a fixed block of seats is leased at a fixed price per seat. We characterize optimal booking limits for both airlines and use the expected revenue function to determine optimal demand and supply curves for seats at a function of price. This enables the design of win-win operational contracts that do not give up control of individual airline networks.

I SD15
Auctions in the Wild: Field Studies of Reverse Auctions
Cluster: Auctions and e-Commerce Invited Session
Chair: Otto Koppius, Assistant Professor of Decision and Information Sciences, RSM / Erasmus University, Department of D&IS (T9-12), PO Box 1738, Rotterdam, 3000 DR, Netherlands, [email protected] 1 - Bid Analyzer: A Method for Price Discovery in Online Reverse Auctions Sandy Jap, Caldwell Research Fellow, Goizueta Business School, Emory University, [email protected]
Online reverse auctions are a pervasive and economically significant marketing activity, generating $40 billion in annual transactional volume in a wide variety of industries. As a result of this, a plethora of point-by-point bid data exists, but there is no systematic approach to analyze it. Hence, we develop a methodology, BidAnalyzer, to estimate the underlying distributions from which bids are generated for each bidder, even if a bidder does not submit a bid in a specific period. This ability to infer the distribution of each bidder is absent in the literature on auctions, and cannot be determined by other time series approaches (e.g., method of moments or vector autoregression). BidAnalyzer also provides the optimal bid price of each bidder, enabling the buyer to understand whether a bidder has bid too aggressively (and may experience a winner’s curse) or has held back (indicating that there is still “money on the table”). We illustrate its practical usefulness using two proprietary datasets and a field experiment. Thus, BidAnalyzer enriches the price discovery efforts of the buyer and provides guidance for improving industrial procurement practice.

2 - Revenue Management for Media Broadcasting Ioana Popescu, Associate Professor of Decision Sciences, INSEAD, Boulevard de Constance, Fontainebleau, 77305, France, [email protected], Victor Araman
Media broadcasting companies seek to allocate limited advertising space across multiple clients and markets in order to maximize profits. We provide simple models and solutions for upfont market allocation among multiple clients under audience uncertainty. We also investigate the dynamic allocation of make-goods for a single client under reversible and irreversible commitment strategies.

2 - The Effects of Auction Design on the Performance of Online Markets for IT Services Eric van Heck, Professor of Information Management and Markets, RSM Erasmus University, 50 Burg Oudlaan, Rotterdam, 3062PA, Netherlands, [email protected], Otto Koppius, Uladzimir Radkevitch
Small business increasingly uses online marketplaces to alocate IT projects to suppliers via reverse auctions. The market and auction design affects auction outcomes and market performance. We draw on electronic markets theory and auction theory to investigate the role of entrance barriers for bidders. Transaction data from two major marketplaces are used to explore barriers’ impact on market performance, i.e. percentage of awarded auctions, percentage of return buyers and buyers’ satisfaction.

3 - Pricing without Market Information Garrett van Ryzin, Columbia University, Uris Hall, New York, NY, 10027, United States, [email protected], Serkan Eren, Costis Maglaras
We look at a variety of classic pricing problems in the case of minimal market information. Using a competitive ratio and maximum regret criteria, we characterize pricing policies that perform well under limited market information and discuss how they relate to classical results in the literature and to industry practices.

4 - Robust Revenue Management Guillaume Roels, Assistant Professor, UCLA Anderson School of Management, 110 Westwood Plaza, Los Angeles, CA, 90077, United States, [email protected], Georgia Perakis
We investigate the problem of allocating fixed capacity in a network to different classes of customers, with limited information about the demand distributions. We consider two criteria for decision-making under uncertainty: maximin and minimax regret. Our approach combines an efficient solution procedure, based on linear programming, with very modest requirements for data about the demand distributions, and is therefore scalable to solve real network revenue management problems.

3 - Efficient Buyer-Seller Relations: Managing Information & Communication Processes in E-RAs Andrea Loesch, PhD Student, Manchester Metropolitan University, Geoffrey Manton Building, Rosamund Street West Off Oxford Road, Manchester, M15 6LL, United Kingdom, [email protected], J. Siân Lambert
Four case studies were conducted in Europe & the US and analyzed using Flanagan’s (1954) Critical Incident Technique. The results illustrate the crucial importance of appropriate inter- & intra-organisational information and communication processes for obtaining efficient buyer-seller relations in ereverse auctions (e-RAs). They also show how important contextual differences between public and private sector buying may affect e-RA applicability. According suggestions managing e-RAs are given.

I SD17
Production & Scheduling I
Contributed Session
Chair: Dan Staley, Oregon State University, 121 Covell Hall, Corvallis, OR, 97331, United States, [email protected] 1 - A Dynamic Programming Approach for Load Control in SingleProduct Tandem Manufacturing Lines Rodrigo Caliz, PhD Student, Penn State University, 1003 West Aaron Drive, Apt. 7B, State College, PA, 16803, United States, [email protected], Jose Ventura
This paper deals with the characterization of the optimal controlled influx rate of a single-product manufacturing line over a finite planning horizon. Deterministic approximations based on both traffic flow theory and non-stationary queueing models are used in order to model the stochastic nature of the system dynamics. Simulation is utilized to evaluate the proposed policy as compared to the noncontrolled system where each job enters the system at its arrival epoch.

4 - Analyzing Bidding Dynamics in Reverse Auctions: Process Metrics and a Functional Approach Otto Koppius, Assistant Professor of Decision and Information Sciences, Rotterdam School of Management, 50 Burg Oudlaan, Rotterdam, Netherlands, [email protected], Wolfgang Jank, Galit Shmueli, Sunil Mithas, Joni Jones

109

SD18

INFORMS PITTSBURGH — 2006
3 - Recent Advances in Geometric Programming Seung-Jean Kim, Consulting Assistant Professor, Information Systems Laboratory, Department of Electrical Engineering, Stanford University, 223 Packard, 350 Serra Mall, Palo Alto, CA, 94305, United States, [email protected], Kwangmoo Koh, Stephen Boyd
Geometric programming has been known since the 1970s. Recently developed solution methods can solve large-scale geometric program (GPs) extremely efficiently and reliably. At the same time, a large number of practical problems in several fields including circuit design and machine learning have been found to be equivalent to (or well approximated by) GPs. We give a brief overview of the new applications of GP and recent advances in solving large-scale GPs.

2 - Forecast Horizon Detection and Solvability of the Concave Cost Production Planning Problem Timothy Lortz, Graduate Research Assistant, University of Michigan, IOE Department, 1205 Beal Avenue, Ann Arbor, MI, 48109, United States, [email protected], Archis Ghate, Robert Smith
We consider a class of production planning problems including the dynamic lotsize model. Under mild assumptions, we show necessary and sufficient conditions for existence of a forecast horizon. We also give a solution algorithm guaranteed to yield an infinite horizon optimal initial decision whenever the problem satisfies these conditions.

3 - A Stochastic Production Planning Problem with Approved Vendor Matrices Ek Peng Chew, Associate Professor, National University of Singapore, 10 Kent Ridge, Singapore, 119260, Singapore, [email protected], Loo Hay Lee, Gang Sun
A disk drive production planning problem with random demands is studied. The manufacturer provides customers with the choice on their preferred suppliers for pairs of inter-dependent components through the approved vendor matrix in order to differentiate its service. We propose a solution approach that provides production plans which minimize the expected total shortage and holding costs while observing the matrix restrictions and limited components supplies.

4 - On Geometric Programming for Solving Nonstochastic Uncertainty Models in Finance Kenneth Kortanek, Visiting Professor, University of Pittsburgh, Department of Industrial Engineering, Pittsburgh, PA, 15261, United States, [email protected]
The underlying law of motion of a financial instrument is a linear differential equation under uncertainty with perturbations for the instrument generating the time series. The dependent variable is an integrand in a discounting mechanism involving an exponentiation, establishing a connection to a two-sided geometric programming problem. Numerical results are presented for the problem of extracting the spot interest rate continuous function from Government Bills, Notes, and Bonds data.

4 - Fuzzy-Logic Based Dynamic Scheduling for Cardboard Manufacturing Company Umut Inan, PhD Candidate, Yildiz Technical University, Department of Industrial Engineering, Huseyin Basl´ gil Doktr y Ogrenci, Istanbul, 34000, Turkey, [email protected], Mesut Yavuz
Today’s fierce competition forces companies to produce numerous end-products, meeting quantities and due dates demanded by the customers. We consider a cardboard packaging company that operates a plant with several key processes in a highly dynamic environment. Currently, many orders have to be processed in a matter of days, resulting in a situation where jobs have to be frequently rescheduled. We develop a fuzzy-logic based dynamic scheduling framework to maximize the efficiency of the plant.

I SD19
Forestry Applications IV
Sponsor: Energy, Natural Resources & The Environment Sponsored Session
Chair: Stephanie Snyder, USDA Forest Service, North Central Research Station, St. Paul, MN, United States, [email protected] 1 - A Simulated Annealing Approach for Treatment Unit Delineation Joseph Petroski, Pennsylvania Department of Conservation and Natural Resources, Bureau of Forestry, Inventory, and Analysis, 137 Penn Nursery Road, Spring Mills, PA, 16875, United States, [email protected], Marc McDill
Spatially explicit timber harvest scheduling requires the delineation of treatment units across a landscape. A multiple objective, simulated annealing model is presented that delineates treatment units using a triangle grid. Treatment units are built by aggregating or splitting groups of adjacent triangles.

5 - Properties of Buffer Allocation in Closed Cyclic Production Systems Dan Staley, Oregon State University, 121 Covell Hall, Corvallis, OR, 97331, United States, [email protected], David Kim
Exetensive simulation testing was conducted on buffer allocation in closed cyclic production systems. Some general rules and behavior have been observed and emprically verified. A comparison to buffer allocation in open systems is made where possible.

I SD18
Geometric Programming: Applications and Software
Cluster: Geometric Programming: Research Inspired by the Work of Dick Duffin, Elmor Peterson & Clarence Zener Invited Session
Chair: Jayant Rajgopal, University of Pittsburgh, Department of Industrial Engineering, 1039 Benedum Hall, Pittsburgh, PA, 15261, United States, [email protected] 1 - Solving Geometric Programing Problems with MOSEK Ulf Worsoe, Senior Developer, MOSEK ApS, Symbion Science Park, Fruebjergvej 3, Box 16, 2100 Copenhagen, Denmark, [email protected]
A geometric programing (GP) problem can be formulated as a convex nonlinear optimisation problem having a special structure, which means that it may be solved to optimality. A number of practical problems, e.g in analog circuit design, can be approximately modelled as geometric programming problems. In this talk we discuss how to formulate and solve large scale geometric programming problems with the software package MOSEK (www.mosek.com), and show some test results.

2 - Green Up and Adjacency Issues in Forest Spatial Harvesting Andres Weintraub, Professor, University of Chile, Department of Industrial Engineering, Santiago, Chile, [email protected], Marcos Goycoolea, Juan Pablo Vielma, Alan Murray
The ARM consists in imposing maximum clearcut size constraints in forest harvest scheduling problems. We discuss green-up extensions of the ARM, distinguishing between static and dynamic green-up constraints. Computational results for different formulations show these are very difficult. In addition, we discuss the impact of different adjacency definitions.

3 - Using Column Generation to Solve Spatially Explicit Forest Management Marc McDill, Associate Professor of Forest Management, Penn State School of Forest Resources, 310 Forest Resources Building, University Park, PA, 16802, United States, [email protected], Babu Rajasekaran
Column generation is a promising method for obtaining near-optimal solutions for large spatially-explicit forest harvest scheduling models. We test a column generation approach with 40 hypothetical problems and 2 methods of generating initial solutions to the sub-problems(seeds). For each method and subproblem, 25, 50 and 100 seeds were generated. Our tests show that both the seed generation method and number of seeds significantly influence solution time and quality of the final solution.

2 - Geometric Programming for Communication Systems Mung Chiang, Assistant Professor, Princeton University, B328 Engineering Quad, Olden Street, Princeton, NJ, 08544, United States, [email protected]
This talk overviews the recent results on how geometric programming provides a powerful tool and a common language to a surprisingly wide range of problems in communication systems, from duality in information theory and performance maximization in queuing theory to wireless network power control and Internet congestion control. It also illustrates how these problems motivate further investigation in success convex approximation by GP and distributed algorithms for GP.

4 - Integer Programming Models for Tactical Harvest and Access Planning Evelyn Richards, Associate Professor, Faculty of Forestry and Environmental Management, University of New Brunswick, Fredericton, NB, E3B 6C2, Canada, [email protected], Eldon Gunn
The problem of scheduling clearcut harvests under adjacency constraints, simultaneous with road construction decisions, is addressed using a mixed integer optimization model. The model allows an unrestricted road network with multiple exit points. Computational results show that it is possible to solve models of practical size using a commercial solver.

110

INFORMS PITTSBURGH— 2006 I SD20
DA Arcade 1
Sponsor: Decision Analysis Society Sponsored Session
Chair: Robin Dillon, Assistant Professor, Georgetown University, 418 Old North, McDonough School of Business, Washington, DC, 20057, United States, [email protected] 1 - A Decision Analysis Approach to Real Option Valuation Ross Shachter, Associate Professor, Management Science and Engineering, Stanford Univeristy, 380 Panama Way, Room 351, Stanford, CA, 94305-4026, United States, [email protected], Stephen Derby
Standard financial option theory assumes rational individuals invest in noarbitrage, equilibrium, complete markets. Our approach values real options without assuming market equilibrium and completeness. Market information constrains real option value without using replicating portfolios. Our method is consistent with the standard theory, therefore conforms to its major results.

SD22

2 - Market Uncertainty and Innovation Incentives in Pharmaceutical R&D John Cavallaro, PhD Candidate, Management Science & Engineering, Stanford University, Stanford, CA, 94305, United States, [email protected], Stefanos Zenios
The newly launched Medicare Part D has renewed both interest in pharmaceutical pricing policies and concerns about innovation incentives. This work quantifies pharmaceutical R&D risk, and then investigates how price reductions affect R&D spending and innovation. Finally, it explores how policy levers can reduce expenditures while maintaining innovation incentives.

3 - Cost-Effectivenessness of Aspirin and Statins for Primary Prevention of CHD Events in Men Stephanie Earnshaw, Head US Health Economics, RTI Health Solutions, 200 Park Offices Drive, Research Triangle Park, NC, 27709, United States, [email protected], Michael Pignone, Jeffrey Tice, Mark Pletcher
A Markov model was developed to examine cost-utility of the effects of aspirin, statin, combination, and no pharmacotherapy for the primary prevention of CHD events. Men at 10-year risk for CHD (2.5% - 25%) were examined. Parameters were taken from literature. Aspirin was more effective and less costly than no treatment in 45-year-old men (7.5% 10-year risk). Addition of statin to aspirin therapy becomes more cost-effective when patient’s 10-year CHD risk is higher than 10%.

2 - Avoiding Disaster Readiness Disasters Robin Dillon, Assistant Professor, Georgetown University, 418 Old North, McDonough School of Business, Washington, DC, 20057, United States, [email protected]
After Hurricane Katrina, people are questioning the readiness decisions that were made. Poor decisions appear to have been made at every level, and tracing the roots of these poor decisions often leads to a discussion of how precursor events shape perceptions. This research examines how past events and currently available probability information are interpreted and examines the importance of integrating decision theory with disaster readiness to improve decision making.

4 - Decision Analytic Methods and Formulary Submissions in Canada Greg Zaric, Professor, Ivey School of Business, University of Western Ontario, London, ON, Canada, [email protected]
A formulary is a list of drugs that will be reimbursed by an insurer. In Canada, the public drug plans in each province require drug manufacturers to submit an economic analysis as part of the formulary submission process. In this talk we discuss the use of cost effectiveness analysis, decision trees, Markov models and other decision analytic techniques that are commonly used by drug manufacturers to produce the required economic analyses.

3 - Combining Expert Judgment Using Structural Equation Models Melissa Kenney, PhD Candidate, Water Quality Modeling and Decision Analysis, Duke University, Nicholas School of the Environment, Box 90328, Durham, NC, 27708, United States, [email protected], Robert Clemen
We describe a method for combining multiple expert judgments using structural equation modeling. The method requires the use of a latent variable that can be viewed as an “ideal” expert judgment, where the individual experts are modeled as multiple measures of the ideal judgment. By using expert data with actual outcomes of the predicted variable and other covariates, we can estimate the relative expert weights and use them in a predictive model. We applied the method to study lake pollution.

I SD22
OR in Information Fusion
Sponsor: Military Applications Sponsored Session
Chair: Rakesh Nagi, Professor, Department of Industrial & Systems Engineering, 438 Bell Hall, University at Buffalo (SUNY), Buffalo, NY, 14260, United States, [email protected] 1 - Maritime Tracking of Past Data Les Servi, Lincoln Laboratory Technical Staff, MIT Lincoln Laboratory, 244 Wood Street, Lexington, MA, 02420-9108, United States, [email protected]
Determining if a ship near a US port is friendly may require inferring the ship’s path during the past week. This can be done using insights from the image processing literature, e.g. Hough Transforms. This talk will present an approach and demonstration, using simulated data, of this capability.

4 - A Review of the Literature on the Concept of Value of Information Lea Deleris, Postdoctoral Researcher, Stanford University, Math Sciences Department, IBM TJ Watson Research Center, Stanford, CA, 94305, United States, [email protected]
The concept of value of information is central to the field of decision analysis: it enables to put bounds on the price to pay for information and can identify decision situations where information is useless. In an age where information is increasingly accessible but often costly, the concept of value of information represents a powerful while simple idea. In this talk, I review the decision analysis literature on the concept of value of information and identify related research directions.

I SD21
Decision Analysis for Healthcare Entities
Sponsor: Decision Analysis Society Sponsored Session
Chair: Anke Richter, Associate Professor, DRMI, Naval Postgraduate School, 699 Dyer Road, Building 234, Monterey, CA, 93943, United States, [email protected] 1 - Strategic Decision-Support for Medical Technology Commercialization Jan B. Pietzsch, President & CEO, Wing Tech Inc., 502 San Benito Avenue, Menlo Park, CA, 94025, United States, [email protected]
The successful commercialization of medical technologies requires informed decision-making and strategy-setting by manufacturers and investors at early stages of the innovation process. This paper presents a novel framework based on systems analysis and probabilistic methods that integrates both engineering-based assessment models and health-economic considerations. The case of a new orthopedic technology is used to demonstrate application of this framework.

2 - Conceptual Spaces as a Framework for Situational Assessment Michael Holender, Student, Center for Multisource Information Fusion, University at Buffalo, Industrial & Systems Engineering, 420 Bell Hall, Buffalo, NY, 14260, United States, [email protected], Moises Sudit, Rakesh Nagi
We consider a geometric representation used in cognitive theory to describe human understanding called Conceptual Spaces. The theory is relevant not only to cognition, but can also handle other types of problems. Our work relates Conceptual Spaces to Data Fusion, namely at Level 2 - Situational Assessment. We introduce mathematical models used to represent Conceptual Spaces and show how they can be useful in considering situations presented by a data fusion system.

3 - Optimizing the Amount and Type of Information at Each Level of Decision-Making Moises Sudit, Managing Director/Professor, University at Buffalo, 414 Bell Hall, Buffalo, NY, 14260, United States, [email protected], Adam Stotz, Rakesh Nagi
As most military decision-makers drown in data and starve for information, issues on the amount and type of data required at each level of decision-making are prevalent and can be viewed as complex optimization problems. We will present Mathematical Models as well as Information Fusion models with corresponding solution strategies that will give the appropriate situation understanding in a sample naval scenario. Extensions of such models will be used for both strategic and tactical operations.

111

SD23

INFORMS PITTSBURGH — 2006
This paper examines the increasing significance of cultural product elements in high-technology markets. Specifically, it uses computer aided verbal and pictorial content analysis to measure changes in the competitive logic of high-technology industries. Findings trace a market preference for cultural components, specifically, product aesthetics, in periods of increased cost based competition and incremental technological innovation.

4 - High Level Data Fusion Using Graph Matching and State Space Search Rakesh Nagi, Professor, Department of Industrial & Systems Engineering, 438 Bell Hall, University at Buffalo (SUNY), Buffalo, NY, 14260, United States, [email protected]
We use attributed graph models to represent situations in Level 2 and 3 fusion (situational awareness and impact assessment). Graph matching is invoked to determine if a situation of interest to the analyst (template graph) exists in a scenario (data graph). A Truncated Search Tree heuristic is developed to perform graph matching.

I SD23
Strategic Decisions in Services Offshoring
Cluster: Global Services Sourcing Invited Session
Chair: Richard Metters, Associate Professor, Emory University, 1300 Clifton Road, Atlanta, GA, 30322, United States, [email protected] 1 - International Outsourcing of Services: A Business Strategy Perspective Stephen Tallman, E. Claiborne Robins Professor of Business, University of RIchmond, Robins School of Business, 1 Gateway Road, Richmond, VA, 23173, United States, [email protected]
Different types of International Outsourcing of Services are considered that might be undertaken by firms and costs and benefits of each are addressed. As firms are increasingly able to separate the physical and the informational aspects of business services and the commodity from the specialized in the information side of services, international outsourcing of commodity information-based services are expected to increase.

2 - Arbitrage as Representation Daniel Beunza, Assistant Professor, Graduate School of Business, Columbia University, Management Division, 708 Uris Hall, 3022 Broadway, New York, NY, 10027, United States, [email protected], David Stark
Existing theories of arbitrage offer a valuable but incomplete understanding of the phenomenon. One literature stream, grounded in neoclassic finance, views arbitrage as a straightforward process of linking like markets for the same security. Another stream, rooted in psychology and behavioural finance, presents arbitrage as an inherently limited in its effects. Both approaches, however, stem from a presentist, under-socialized and over-abstracted view of individual calculation that ignores Knightian uncertainty. To address this shortcoming, we undertake a three-year ethnographic study of arbitrage trading at the trading room of a Wall Street investment bank. We find that arbitrage is a cogitative and material process that emerges from the juxtaposition of information, calculative frames, visualization and social relations. In contrast to the existing views of arbitrage as perfect linking device or flawed repository for herding, our view suggests that arbitrageurs constitute a forum for resolving differences in beliefs about the economy.

3 - Market Watch: Information and Availability Cascades in the U.S. IPO Market Tim Pollock, Associate Professor of Management, Smeal College of Business, Pennsylvania State University, 417 Business Building, University Park, PA, 16802, United States, [email protected], Patrick Maggitti, Violina Rindova
In this study we advance current research on the dynamics of social influence in markets by examining how information and availability cascades influence the attentional and evaluative choices that the media and investors make about newly public firms. We integrate emerging theoretical perspectives on information and availability cascades by examining the extent to which the media and investors follow the logics of each type of cascade in deciding which firms to focus their attention on and how to evaluate them. We extend current research on cascades by positing that the choices of market participants reflect both intra-cascade effects arising from imitation of others in one’s own professional community, and inter-cascade effects arising from efforts to glean information from the choices of others in another relevant professional community.

2 - Host Country Efficiency in Services Offshoring: A Comparative Approach Eugene Hahn, Assistant Professor, Department of Information & Decision Sciences, Salisbury University, Salisbury, MD, 28101, United States, [email protected], Jonathan Doh, Kraiwinee Bunyaratavej
Service offshoring is an increasingly important aspect of the global economy. This phenomenon’s growth has lead to increased interest in services offshoring from a diverse set of groups. Yet to date empirical research in services offshoring has been somewhat limited. In this study we empirically examine services offshoring capabilities of countries and compare them utilizing an efficiency-based perspective. Implications for management decision makers and governmental officials are discussed.

I SD25
Select Business Modeling Case Studies in Telecommunications Industry
Cluster: OR Practice Invited Session
Chair: Narayan Raman, Lucent Technologies, 101 Crawfords Corner Road, Holmdel, NJ, 07733, United States, [email protected] 1 - Analysis of Cellular and Wireline Voice Convergence in the Consumer Market Abdol Saleh, Lucent Technologies, Business Modeling Department, 101 Crawfords Corner Road, Holmdel, NJ, 07733, United States, [email protected], Yuliy Baryshnikov
To provide converged Cellular/Wireline services, telecom operators seek clear economic drivers to develop service offer strategies. We study the impact of factors, such as market segment characteristics, price sensitivity, network investment, service bundling, operation cost and service differentiation in an environment where various operators adopt different go-to-market strategies. We quantify the impact of these strategies on market share acquisition and churn reduction for each operator.

3 - A Typology of Offshoring and Outsourcing for Electronically Transmitted Services Richard Metters, Associate Professor, Emory University, 1300 Clifton Road, Atlanta, GA, 30322, United States, [email protected]
A normative model of the appropriate role of offshoring is proposed. We present a strategic contingency model, to be viewed at the process level, intimating that firms with the same processes should come to different solutions regarding the offshoring and outsourcing decisions.

I SD24
Language, Culture, and Representation in Markets
Sponsor: Organization Science Sponsored Session
Chair: Amit Nigam, Post Doctoral Researcher, Rotman School of Management, University of Toronto, 105 St. George Street, Toronto, ON, M5S 3E6, Canada, [email protected] Co-Chair: Klaus Weber, Assistant Professor of Management and Organizations, Kellogg School of Management, Northwestern University, 2001 Sheridan Rd, Evanston, IL, 60208-2001, United States, [email protected]
The papers in this session examine role of language, information and representational practices in core market processes.

2 - Business-Modeling Framework for Radio Access Network Capacity Planning in CDMA Networks Alina Ionescu-Graff, Lucent Technologies/ Bell Labs, 101 Crawfords Corner Road, 4K-403, Holmdel, NJ, 07733, United States, [email protected], Doru Calin, Ednny Mari Aguilar
A Cellular Capacity Planning (CPP) service for Radio Access Networks was developed to help Service Providers plan for cellular networks to meet Service Level Agreements but without over provisioning. This paper presents a businessmodeling framework to quantify the cost/benefits of the CCP Service by translating the provisioning error reductions into capital and operations savings, and revenue preservation and increase (from reduction of excess equipment, blocking, churn, customer care ops, etc)

1 - Language as a Window to Cultural and Technological Markets Micki Eisenman, Assistant Professor, Zicklin School of Business, Baruch College, CUNY, Management Department, Box B9-240, New York, NY, 10026, United States, [email protected]

112

INFORMS PITTSBURGH— 2006
3 - Economics of Hosted Applications Narayan Raman, Lucent Technologies, 101 Crawfords Corner Road, Holmdel, NJ, 07733, United States, [email protected]
This study evaluates the benefit to a network operator of outsourcing the delivery of applications to consumer or enterprise end users. It investigates the impact of market- and supply-related uncertainties on the relative economics of the outsourcing decision.

SD28

Richard A. Davis
This paper considers the problem of detecting break points for a broad class of non-stationary time series models. In this formulation, the number and locations of the break points are assumed unknown. Each piece is assumed to be stationary and can be modeled from a class of parametric time series models. The minimum description length is used as a criterion for estimating the number of break points, the location of break points, and the parametric model in each segment.

I SD26
Mathematical Programming in Data Mining Machine Learning
Sponsor: Data Mining Sponsored Session
Chair: Alkis Vazacopoulos, Director, Dash Optimization Inc., 560 Sylvan Avenue, Englewood Cliffs, NJ, 07632, United States, [email protected] 1 - Consistent Biclustering via Fractional 0-1 Programming Stanislav Busygin, PhD Student, Department of Industrial and Systems Engineering, University of Florida, 303 Weil Hall, Gainesville, FL, 32611, United States, [email protected], Oleg Prokopyev, Panos Pardalos
Biclustering consists in simultaneous partitioning of the set of samples and their features into classes. Samples and features classified together are supposed to have a high relevance to each other. We define the notion of consistent biclustering and prove that it implies separability of the classes by convex cones. We discuss both supervised and unsupervised learning algorithms. The developed models involve fractional 0-1 optimization. Computational results on microarray data are reported.

2 - Selecting the Best System When Systems are Revealed Sequentially Barry Nelson, Professor, Northwestern University, Department of IEMS, 2145 Sheridan Road, Evanston, IL, 60208-3119, United States, [email protected], L. Jeff Hong
Statistical ranking and selection (R&S) is a collection of experiment design and analysis techniques for selecting the system with the largest or smallest mean performance. The existing procedures all assume that the set of alternatives is available at the beginning of the experiment. In many situations, however, the alternatives are revealed sequentially during the experiment. We introduce new procedures that are capable of selecting the best alternative in these situations.

3 - Enhancements of the Generalized Likelihood Ratio Tests for On-Line Detection of Changes Emmanuel Yashchin, Research Staff Member, IBM Research, T.J. Watson Research Center, Room 33-212, Box 218, Yorktown Heights, NY, 10598, United States, [email protected]
We discuss the problem of monitoring in the presence of abrupt changes, such as shifts or drifts. We introduce a unified methodology based on use of likelihood ratio tests that enables one to obtain control schemes that provide both good statistical performance and are relatively easy to implement in the sense that they depend on very few design parameters and require a limited computational effort that is dynamically adjusted based on process conditions.

2 - Network-Based Approaches in Data Mining Sergiy Butenko, Assistant Professor, Texas A&M University, College Station, TX, 77843, United States, [email protected], Svyatoslav Trukhanov, Balabhaskar Balasundaram
This talk will discuss data mining techniques based on network representations of data sets. In particular, alternative ways of representing the data as a network will be described, several models describing structural properties of the network will be introduced, and solution methods for the corresponding optimization problems will be described.

4 - Bayesian Sequential Detection and Isolation of an Unobservable Change Savas Dayanik, Assistant Professor, Princeton University, ORFE, EQuad, Princeton, NJ, 08544, United States, [email protected], Christian Goulding, H. Vincent Poor
Suppose that the characteristics of a stochastic process change suddenly at an unobservable time to one of several alternatives. The problem is to detect simultaneously the change time and change type as quickly as possible after change happens. In a Bayesian setup we describe an optimal sequential decision rule and illustrate it on numerical examples.

3 - Pseudo-Boolean Regression Peter L. Hammer, Director and Professor, Rutgers University RUTCOR, 640 Bartholomew Road, Piscataway, NJ, 08854-8003, United States, [email protected], Tibérius Bonates
Generalizing the LAD (Logical Analysis of Data) approach, an iterative procedure is proposed to construct a sequence of polynomials of binary (0-1) variables, approximating a real-valued target function with prescribed values on a set of binary vectors. Solutions with correlations exceeding 97-98% are efficiently constructed on all benchmark problems considered.

I SD28
Panel Discussion: Teaching Experimental Design and Statistical Process Control Methods
Sponsor: Quality, Statistics and Reliability Sponsored Session
Chair: Jye-Chyi (JC) Lu, Professor, Georgia Institute of Technology, 765 Ferst Drive, Atlanta, GA, 30022, United States, [email protected] Co-Chair: Dan Apley, Associate Professor, Northwestern University, IE/MS, 2145 Sheridan Road, Evanston, IL, 60208, United States, [email protected] 1 - Panel Discussion: Teaching Experimental Design and Statistical Process Control Methods in IE/OR/Statistics Programs Moderator: Jeff C. F. Wu, Professor, Georgia Institute of Technology, Industrial and Systems Engineering, 765 Ferst Drive, Campus Box 0205, Atlanta, GA, 30332, United States, [email protected], Panelists: Bruce Ankenman, Jye-Chyi (JC) Lu, Russell Barton, George Runger, Yu Ding, Dan Apley
In this session a panel of faculty who teach undergraduate and graduate level courses in design of experiment (DOE) come together to present and exchange information about their courses, in terms of the course design, topics covered, data supplied, and software and text books used. This session is intended for anyone who is currently teaching or hoping to create a course on the subject of DOE.

4 - Clustering by QUBO (Quadratic Unconstrained Binary Optimization) Gabriel Tavares, Dash Optimization Inc., 560 Sylvan Avenue, Englewood Cliffs, NJ, 07632, United States, [email protected], Peter L. Hammer, Endre Boros
Numerous algorithmic graph theory problems (MAX-CLIQUE, MIN-COV, MAXCUT, BALANCING) are formulated as quadratic unconstrained binary optimization problems, allowing the exact solution within seconds of real life problems (coming from telecommunications, VLSI design, biological/social networks) with 500,000 (and more) vertices. Applications to clustering problems with millions of objects are presented.

I SD27
Sequential Analysis and Model Selection Problems
Sponsor: Quality, Statistics and Reliability Sponsored Session
Chair: Savas Dayanik, Assistant Professor, Princeton University, ORFE, E-Quad, Princeton, NJ, 08544, United States, [email protected] 1 - Structural Break Estimation for Non-Stationary Time Series Signals Thomas Lee, Associate Professor, Colorado State University, Department of Statistics, Fort Collins, CO, 80523, United States, [email protected], Gabriel Rodriguez-Yam,

113

SD29 I SD29

INFORMS PITTSBURGH — 2006
3 - An MDP Model for Optimizing Statin Start Times for Diabetes Patients Murat Kurt, Doctoral Candidate, University of Pittsburgh, 4135 Murray Avenue, Pittsburgh, PA, 15217, United States, [email protected], Nilay D. Shah, Brian Denton, Sandra C. Bryant, Steven A. Smith
Diabetes affects approximately 7% of the U.S. population and costs more than $130 billion annually. Several risk models are used by clinicians to guide diabetes related treatment decisions.We examine an important treatment decision, the optimal start time of statins, using a Markov decision process (MDP) model. Our MDP model indicates that significant discrepancies among alternative risk models translate into significant differences in the implied optimal time to start treatment with statins.

System Analysis and Design in Healthcare
Cluster: Joint Cluster Healthcare/ HAS: Healthcare Engineering Invited Session
Chair: Yuehwern Yih, Professor of Industrial Engineering, Purdue University, 315 N. Grant Street, W. Lafayette, IN, 47907, United States, [email protected] 1 - Specialization and Competition in Healthcare Delivery Networks Vikram Tiwari, Indiana University, Kelley School of Business, Operations & Decision Technology, Bloomington, IN, 47405, United States, [email protected], H. Sebastian Heese
Hospital networks that offer multiple services at multiple locations can improve cost and quality by creating specialized facilities; however, this may be at the cost of market share. Using a spatial model, we study when specialization is profitable, and which facilities and services provide the greatest value through specialization.

2 - Holiday Operating Room On-Call Staffing Requirements Franklin Dexter, Professor, University of Iowa, Department of Anesthesia, 6-JCP, Iowa City, IA, 52242, United States, [email protected]
The numbers of cases starting during each holiday period is a statistically valid end-point for operating room managers to use when evaluating how busy are holidays relative to weekend days. The statistic is useful when combined with mathematically valid assessments of appropriate weekend staffing. The latter are set covering problems.

4 - Maximizing the Efficiency of the U.S. Liver Allocation System through Region Selection (Submitted to for the 2006 Pierskalla Award) Nan Kong, Assistant Professor, Industrial and Management Systems Engineering, University of South Florida, ENC 2505, Tampa, FL, 33620, United States, [email protected], Brady Hunsaker, Andrew Schaefer, Mark Roberts
Allocating organs for transplantation has been a contentious issue in the United States for decades. One of the many remaining questions is how to cluster Organ Procurement Organizations (OPOs) into regions such that allocation efficiency at the regional level is maximized. We formulate the above problem as a large-scale set-partitioning problem and adapt branch and price to solve it. Computational results are presented in the talk.

3 - A Structural Equation Model for Clinicians’ Adherence to Clinical Practice Guidelines Sze-Jung Wu, Research Assistant, Purdue University, School of Industrial Engineering, 315 N. Grant Street, W. Lafayette, IN, 47907, United States, [email protected], Mark Lehto, Yuehwern Yih, Mindy Flanagan, Alan Zillich, Brad Boebbeling
Clinical practice guidelines (CPG) offer clinicians instructions such as which diagnostic or screening tests to order. Previous studies reported that CPG could improve health outcomes, but had limited effect on changing physicians’ behavior. In order to study clinicians’ adherence to CPG, we conducted a survey instrument, including clinicians’ background, perceptions, and actions on CPG. Structural equation modeling was applied to validate the feasibility of our proposed conceptual model.

I SD31
Optimization Approaches in Financial Engineering II
Cluster: Financial Engineering and Risk Management Invited Session
Chair: Stan Uryasev, Professor, University of Florida, Department of Industrial and Systems Engineering, Gainesville, FL, 32601, United States, [email protected] 1 - Valuation of Electricity Generation Facilities: Optimal Exercise Boundaries Approach Valeriy Ryabchenko, PhD Student, University of Florida, 303 Weil Hall, PO Box 116595, Gainesville, FL, 32611-6595, United States, [email protected], Stan Uryasev
The paper presents an optimization approach to valuation of electricity generation assets. We build a LP model to find the optimal operation boundaries for an electricity generation plant. We consider different operational constraints: start-up and shut-down costs, ramp-up time, output dependent operating heat rate. The model is furnished with a set of monotonicity constraints on decision variables. We test the model with historical and simulated sample paths.

I SD30
Optimization of Medical Decisions
Sponsor: Health Applications Section Sponsored Session
Chair: Brian Denton, Assistant Professor, Mayo Clinic College of Medicine, 200 First Street SW, Rochester, MN, 55906, United States, [email protected] 1 - Pattern Recognition and Classification in Medical Diagnosis Eva Lee, Associate Professor, School of Industrial and Systems Engineering, Georgia Institute of Technology, Atlanta, GA, 30332, [email protected]
The cardiovascular system provides oxygen/nutrients to the body. Tissue ischemia, due to obstruction of arterial blood flow, can result in serious effects (vision/limb loss). Development of new vessels can fuel the progression of cancer and macular degeneration. We present a pattern recognition algorithm and an optimization-based predictive model to analyze artery patterns. We illustrate its use for disease diagnosis/monitoring via classification of healthy vs diseased microvascular networks.

2 - Pricing CDO Tranches Using Reduced-Form Default Model Alex Nakonechnyi, PhD Student, University of Florida, 303 Weil Hall, PO Box 116595, Gainesville, FL, 32611-6595, United States, [email protected]
This research presents a framework for pricing Collateralized Debt Obligation tranches. It applies maximum entropy principle to estimating default distribution implied by the market prices (spreads) of CDO tranches. This default distribution is used to price other credit derivatives.

3 - Efficient Execution in Secondary Mortgage Market Chung-Jui Wang, PhD Student, University of Florida, 303 Weil Hall, PO Box 116595, Gainesville, FL, 32611-6595, United States, [email protected], Stan Uryasev
This paper performs secondary mortgage marketing analysis which considers functionality including the loan-level execution, guarantee fee buy-up/buy-down option, servicing retain/release option, and excess servicing fee. Retaining servicing or excess servicing fee involves uncertainty cash flows due to interest rate fluctuation and prepayment. We consider conditional value at risk (CVaR) as risk measure and perform a mean-CVaR efficient execution.

2 - On the Nearest Dynamic Time Warping Neighbor for Abnormal Brain Activity Classification W. Art Chaovalitwongse, Assistant Professor, Rutgers University, 96 Frelinghuysen Road, Piscataway, NJ, 08854, United States, [email protected], Rajesh Sachdeo, Ya-Ju Fan
Uncontrolled epilepsy poses a significant burden to society due to associated healthcare cost to treat and control the unpredictable and spontaneous occurrence of seizures. In this talk, we present a novel classification technique, Nearest Dynamic Time Warping Neighbor, used to classify normal and epileptic brain activities through quantitative analyses of electroencephalogram (EEG) recordings. This technique employs dynamic programming to excavate hidden patterns/relationships in the brain.

114

INFORMS PITTSBURGH— 2006 I SD32
Corporate Financial Modeling
Sponsor: Financial Services Sponsored Session
Chair: Peter Meindl, PhD Candidate, Stanford University, 1038 Guinda Street, Palo Alto, CA, 94301, United States, [email protected] 1 - Constrained Index Tracking Using Stochastic Receding Horizon Control (SRHC) Chang Hwan Sung, PhD Candidate, Stanford University, Terman Engineering Center, Room 479, Stanford, CA, 94305, United States, [email protected], James Primbs
We use SRHC as a suboptimal approach to an infinite horizon constrained index tracking problem. The dynamics are lifted to handle constraints that involve state and control information at different points in time, and the finite horizon optimizations in RHC can be written as an SDP under the structure of innovations feedback. Computational results using the historical data show that SRHC is a promising approach in solving difficult portfolio optimization problems with constraints.

SD34

3 - Determining the Number of Slots to Auction David Lovell, Associate Professor, University of Maryland, Department of Civil Engineering, 1173 Martin Hall, College Park, MD, 20742, United States, [email protected], Avijit Mukherjee, Michael Ball, Andrew Churchill
We propose an integer programming model to determine the number of slots to auction during different hours of the day at an airport, to maximize the net benefit to airlines from scheduling operations. Scheduling an operation during a given hour generates revenues for airlines, while delaying and canceling causes them financial loss. The net value can be determined by subtracting the cost of flight delays and cancellations from the overall revenue that is generated from scheduling operations.

4 - Fair Allocation of Slot Leases Michael Ball, Professor, Robert H. Smith School of Business, University of Maryland, College Park, MD, 20742, United States, [email protected], Ming Zhong
Under certain proposals for airport slot controls, slot access rights have limited terms. To ease future transitions, these terms should be staggered. We study the problem of creating a fair allocation such slot “leases” where fairness is measured against historical usage.

2 - Corporate Bond Portfolio Optimization with Transaction Costs Peter Meindl, PhD Candidate, Stanford University, 1038 Guinda Street, Palo Alto, CA, 94301, United States, [email protected], James Primbs
We propose a methodology for corporate bond portfolio optimization in a multiperiod environment with transaction costs. Our methodology melds stochastic programming with receding horizon control, essentially breaking down the portfolio optimization problem into a sequence of problems solved over time. Through Monte Carlo simulation of reduced form defaultable bonds, we demonstrate results showing our methodology can significantly outperform other methodologies.

I SD34
Distribution Problems in Facility Logistics I
Sponsor: Transportation Science & Logistics Sponsored Session
Chair: Andy Johnson, Assistant Professor, Department of Industrial and Systems Engineering, Texas A&M University, 241 Zachry, 3131 TAMU, College Station, TX, 77843, United States, [email protected] 1 - Airport Terminal Design Using Simulation Cenk Tunasar, PhD, Senior Associate, Booz Allen Hamilton, 8283 Greensboro Drive, McLean, VA, 22102, United States, [email protected], Rahmi Evlioglu
We demonstrate the use of simulation in airport terminal facility design via examples from JFK, Baltimore-Washington and Atlanta Airports. The models, developed by languages Arena and Automod, cover detailed movements of passengers and baggage. Working with architects and planners, we show that simulation can become part of the design process by evaluating multiple design options quickly. We will also share the challenges we faced and lessons learned in implementing simulation modeling.

3 - A Semidefinite Programming-Based Receding Horizon Control Approach to Portfolio Optimization James Primbs, Assistant Professor, Stanford University, 444 Terman Engineering Center, Stanford, CA, 94305, United States, [email protected]
We present a new approach to portfolio optimization based on the receding horizon control methodology. The approach involves the solution of on-line optimizations in the form of semidefinite programs that are able to incorporate various forms of constraints. Overall, it has attractive computational features and represents a promising new approach to portfolio optimization.

I SD33
Air Traffic Management
Sponsor: Aviation Applications Sponsored Session
Chair: Michael Ball, Professor, Robert H. Smith School of Business, University of Maryland, College Park, MD, 20742, United States, [email protected] 1 - Scenario-Free Approaches to the Stochastic Ground Holding Problem Mark Hansen, Professor, University of California at Berkeley, Civil and Environmental Engineering, 114 McLaughlin Hall, Berkeley, CA, 94720, United States, [email protected], P. Barry Liu
We investigate methods for solving the Stochastic Ground Holding Problem that do not require the use of a small set of capacity scenarios. The advantage of this approach that it better reflects actual capacity profiles, which evolve stochastically in a manner that is only roughly approximated by a small set of scenarios. The disadvantage is that the “curse of dimensionality” must be overcome. We present methods for doing this, and assess their performance.

2 - A Simulated Annealing-Based Heuristic for Outbound Door Assignment in LTL Crossdock Terminals Hector Carlo, Instructor at University of Puerto Rico, PhD Candidate, University of Michigan, 1205 Beal Avenue, Ann Arbor, MI, 48109-2117, United States, [email protected], Yavuz Bozer
We present a simulated annealing-based heuristic to obtain outbound trailer-todoor assignments that minimize the overall material handling effort in a crossdock over a specified time period. The heuristic was used to determine door assignments in a major crossdock facility operated by our corporate collaborator.

3 - Design Analysis of a New Class of Order-Picking Systems Margarit Khachatryan, Graduate Research Assistant, School of Industrial and Systems Engineering, Georgia Institute of Technology, 765 Ferst Drive, NW, Atlanta, GA, 30332, United States, [email protected], Leon McGinnis
We develop a formal model for the design analysis of a new class of order-picking systems: small parts high volume order picking with pick-to-buffer. Viewing the process as two stages, order picking and order assembly, leads to a tandem queuing model of performance. Analyzing the model requires new results for both picker travel time, the estimation of the squared coefficient of variation for departures from a queue with batch arrivals and model for the second stage: order assembly.

2 - Command and Control Modeling for Air Traffic Flow Management Robert Hoffman, Principal Analyst, Metron Aviation, 131 Elden Street, Herndon, VA, 20170, United States, [email protected]
In an effort to cope with air traffic demand growths and changing demand patterns, government and industry have developed simulation tools to model future concepts of air traffic management. We present a model for system-wide traffic flow management, which can be attached to existing simulation tools as a module. The model is multi-objective, considering efficiency, equity, and flexibility. Integer programs and network flow formulations are used to capture efficiency in the model.

4 - Benchmarking Warehouse Performance Andy Johnson, Assistant Professor, Department of Industrial and Systems Engineering, Texas A&M University, 241 Zachry, 3131 TAMU, College Station, TX, 77843, United States, [email protected], Leon McGinnis
The Internet-based Data Envelopment Analysis System (iDEAs) for System-based Self-assessment of Warehouse Operations is a tool available via the Internet. It allows a user to compare the performance of their warehouse to a database of warehouses in order to understand their current performance level. This presentation will describe the most recent results of a review of 390 warehouses, the trends identified, the relationships between performance and a variety of practices and attributes.

115

SD35 I SD35

INFORMS PITTSBURGH — 2006
2 - A Multi-Product Network Design Model with Lead Time and Safety Stock Considerations Karthik Sourirajan, Research Staff Member, IBM TJ Watson Research Center, 1101 Kitchawan Road, Yorktown Heights, NY, 10598, United States, [email protected], Reha Uzsoy, Leyla Ozsen
We study a multi-product network design problem in which we explicitly model the non-linear relationships between demand assigned to a DC, replenishment lead times and safety stocks. While lead time congestion can be thought of as a global network property that depends on the total demand flows, safety stock risk-pooling can be thought of as a local network property that depends on individual products’ demand flows. Our model captures the trade-off between these properties.

Network Design and Other Strategic Issues in Transportation
Sponsor: Transportation Science & Logistics Sponsored Session
Chair: Ahmad Jarrah, Assistant Professor of Decision Science, University of Wyoming, College of Business Building 216, Laramie, WY, 82071, United States, [email protected] 1 - Service Based Approach for Supply Chains with Multiple Transportation Modes Carlos Eloir Blanco Rodriguez, Universidad Autonoma de Nuevo Leon, PISIS/FIME, Av. Universidad s/n., Ciudad Universitaria AP. 076’F’, San Nicolas de los Garza, NL, 66450, Mexico, [email protected], Deniz Ozdemir
We consider a supply network management problem with multiple products, where periodic demand of nonidentical stocking locations are satisfied via multiple transportation modes. We incorporate the unexpected delays in Marine terminals due to increased traffic. We formulate the problem as MIP to find an optimal assignment of routes and transportation modes. We extend the problem formulation to stochastic case. The impact of terminal congestion is demonstrated for different scenarios.

3 - Two Product Inventory Control Systems with Substitution Olga Ortiz, PhD Student, Georgia Institute of Technology, 765 Ferst Drive NW, Atlanta, GA, 30332, United States, [email protected], Alan Erera, Chip White
Consider a two-product inventory system where demand is observed only through sales. When a stock-out occurs, substitution can take place further reducing demand observability. We present approaches based on partially observed Markov decision processes to assess the potential benefit from improving demand observability in this setting.

2 - Finding the Minimum Network that Maximizes Service Hui Chen, [email protected], Ann Campbell, Barrett Thomas
Many national delivery providers offer delivery time commitments between cities. These commitments direct the design of the delivery network. In this network, each arc defines a direct connection between cities and represents a substantial investment. The minimum cost network that connects all cities is a tree, but a tree may not satisfy all of the delivery commitments. We find a tree that minimizes the maximum delivery time violation and present computational results based on the U.S.

4 - Approximate Dynamic Programming for Inventory Allocation Under Uncertainty Huseyin Topaloglu, Assistant Professor, Cornell University, 223 Rhodes Hall, Ithaca, NY, 14853, United States, [email protected], Sumit Kunnumkal
We present two approximate DP approaches for an inventory allocation problem. The first one uses linear approximations of the value function and exploits the LP formulation of the underlying DP. The second one uses Lagragian relaxation to relax the constraints that link the decisions for different production plants.

3 - Allocating Costs of a Collaboration in Transportation Procurement in a Multi Carrier Environment Okan Orsan Ozener, Graduate Research Assistant, H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology, 765 Ferst Drive NW, Atlanta, GA, 30332, United States, [email protected], Özlem Ergun
Collaboration improves efficiency along-or-across supply chains by exploiting synergies between shippers and carriers. We study a network where shippers collaborate to minimize system wide repositioning costs by negotiating prices with several carriers. The problem of finding the optimal routes that minimize total cost of covering all the shippers’ demand is NP-Hard. We develop an algorithm to solve this optimization problem that uses integer programming as a subroutine.

I SD37
Economics of IS 2
Sponsor: Information Systems Society Sponsored Session
Chair: Vidyanand Choudhary, Assistant Professor, University of California, Irvine, Paul Merage School of Business, SB 313, Irvine, CA, 92697, United States, [email protected] 1 - Allocating Objects in a Network of Caches: Centralized and Decentralized Analyses Karthik Kannan, Assistant Professor, Purdue University, 403 West State Street, West Lafayette, IN, 47907, United States, [email protected], Mohit Tawarmalani, Prabuddha De
In this paper, we analyze object allocation in a network of caches that share web content to exploit network externality benefits. The analysis is presented for both centralized and decentralized scenarios, and is carried out using operations research and game-theoretic tools. The optimal allocation is found for each case, and cache incentives are aligned with the socially optimal welfare by devising appropriate pricing mechanisms.

4 - Large-Scale Less-than-Truckload Service Network Design Ahmad Jarrah, Assistant Professor of Decision Science, University of Wyoming, College of Business Building 216, Laramie, WY, 82071, United States, [email protected], Ellis Johnson
We present a novel formulation for designing large less-than-truckload (LTL) freight networks. The model captures the network design constraints, the load plan’s tree structure and other LTL-specific requirements. We fragment the massive model with up to 2 million 0-1 variables and 2 million rows, into a separate and efficient IP for each destination/ due day-of-week combination along with a master problem. Our results yield major annual savings (~ $20 million).

I SD36
Inventory, Production and Distribution Applications
Sponsor: Transportation Science & Logistics Sponsored Session
Chair: Huseyin Topaloglu, Assistant Professor, Cornell University, 223 Rhodes Hall, Ithaca, NY, 14853, United States, [email protected] 1 - Integrated Scheduling of Production and Distribution with Capacited Vehicles and Production Set-ups Su Gao, PhD Candidate, MSIS Department, Rutgers University, PhD Program #76, 180 University Avenue, Newark, NJ, 07102, United States, [email protected], Ronald Armstrong, Lei Lei
Coordination of production and distribution is an important issue faced by many industries. We develop a cyclic scheduling model with capacited trucks and products with short-life, set-ups between productions and non-instantaneous routing. We analyze optimal properties of special cases and propose an effective approximation algorithm for the general problem. The objective is to minimize the cycle time subject to the customer demand, product lifespan and capacity constraints.

2 - Electronic Commerce and Local Competition Anindya Ghose, Assistant Professor, Stern School of Business, New York University, 44 West Fourth Street, KMC 8-94, New York, NY, 10012, United States, [email protected], Avi Goldfarb, Chris Forman
We examine how changes in the local supply of goods in the offline world changes consumer purchase behavior online. The emergence of new online retailing channels may act as a substitute for the benefits of urban concentration both by offering lower prices and by providing increased product variety. We explore these questions using data from Amazon on the top twenty books, and DVDs for over 8626 unique locations over 10 months. We discuss implications for the Long Tail.

3 - Performance Based Advertising Jane Feng, Assistant Professor, University of Florida, Gainesville, FL, 32611, United States, [email protected]
Performance based pricing, where the advertisers only pay to the publisher when the advertisement generates valid results, becomes more and more popular in the internet advertising industry. How does this pricing scheme impact the publisher’s quality, the advertisers’ welfare, compared to the traditional impression based advertising? This paper tries to understand its basic economics.

116

INFORMS PITTSBURGH— 2006 I SD38
Roundtable II - Tactical and Operational Capacity Management
Sponsor: Railroad Applications Sponsored Session
Chair: Dharma Acharya, AVP - Operations Research, CSX Transportation, Jacksonville, FL, 32202, United States, [email protected] 1 - Revolutionary Functionality via Evolutionary Development Ron Lindsey, Owner, Communication Architecture, 1828 Cherry Street, Jacksonville, FL, 32205, United States, [email protected]
A railroad’s traffic capacity is constrained by technologies that were introduced in the first 1st and 2nd quarters of the last century, i.e., track circuits and voice radio respectively. With more timely train position and speed data available via wireless data, railroads can project traffic conflicts and re-plan accordingly, thereby increasing capacity. Such capability can be done with minimal additional infrastructure and with out taking on a railroad’s current CAD platform.

SD40

2 - Some Thoughts on Getting from Here to ‘There’ Marshall Fisher, UPS Professor, The Wharton School, University of Pennsylvania, Philadelphia, PA, 19104, United States, [email protected]
I’ll react to earlier speakers suggestions of where we should be going and offer some specific suggested action steps of how to get there.

3 - How to Create Change within the Camelot? Christopher Tang, Professor, UCLA Anderson School, 110 Westwood Plaza, Los Angeles, CA, 90095, United States, [email protected]
The world is changing, and OR/MS must change along with it. Previous sessions have examined pertinent developments and emerging opportunities and challenges for OR/MS practitioners, researchers, and educators. This session discusses possible concerted actions to promote faster progress in some of these areas.

I SD40
Recent Developments in Location Analysis
Sponsor: Location Analysis Sponsored Session
Chair: Robert Aboolian, Assistant Professor, California State University at San Marcos, College of Business Administration, 333 S. Twin Oaks Valley Road, San Marcos, CA, 92096, United States, [email protected] 1 - Location-Allocation of Service Units on a Congested Network Oded Berman, Professor, Rotman School of Management, University of Toronto, 105 St. George Street, Toronto, ON, M5S 3E6, Canada, [email protected], Zvi Drezner, Robert Aboolian
We consider the problem of locating and allocating servers on a congested network. The service and demand processes are assumed to be Poisson distributed. The objective is to minimize the total fixed installation, variable, travel time and waiting time costs. Analysis of the problem, exact and approximate solution approaches are presented.

2 - Factors Affecting Railroad Yard Performance Carl Martland, Senior Research Associate, Massachusetts Institute of Technology, Room 1-153, Cambridge, MA, 02139, United States, [email protected]
The average time spent in rail freight classification yards is commonly more than 30 hours, more than 50% over the benchmarks achieved at various times and locations in the past three decades. Long yard times result primarily from strategic decisions concerning yard capacity and service design that are based in part upon perceptions of equipment, yard, and train costs. Better understanding of these costs - and of the potential for delays - could lead to a renewed concern for yard performance.

3 - Lean Production Applications at CPR Freight Classification Yards Jeff Adams, General Manager - Yards Operations, Canadian Pacific Railway, Suite 500 Gulf Canada Square, 401 9th Avenue SW, Calgary, AB, T2P 4Z4, Canada, [email protected]
Addressing railroad capacity constraints often leads to expanded infrastructure or other hard asset investments - an extremely expensive way to buy capacity. Lean production methodologies can yield significant results without resorting to capital. Experiences and results of lean applications in freight classification yards at Canadian Pacific will be discussed.

2 - Lost Demand in a Competitive Environment Zvi Drezner, Professor, California State University, Fullerton, CA, 92834, United States, [email protected], Tammy Drezner
In many competitive situations not all the buying power is spent at the competing facilities because there are substitute products available. These substitute products are not as desirable as the product in question, but customers will purchase the less desirable substitutes if they are more conveniently available. We construct a model that considers a decline in demand as a function of the distance to competing facilities. Solution methods are proposed and computational experiments reported.

4 - Improving Railroad Classification Terminal Performance Using Concepts of “Lean Railroading” Jeremiah Dirnberger, Specialist, Yard Operations Performance, Canadian Pacific Railway, 401 9th Avenue SW, Suite 500, Calgary, AB, T2P4Z4, Canada, [email protected], Christopher Barkan
The high potential profitability of carload traffic suggests that railroads further grow this type of traffic by providing more reliable service. The classification terminal is a key determinant of service reliability. Performance improvement can be gained by adapting tools from manufacturing. This new approach, introduced here as “Lean Railroading,” comprises the adaptation of Lean, the Theory of Constraints (TOC) and Statistical Process Control (SPC or “six sigma”) to the terminal system.

3 - Network Location with Uniform Pricing Dmitry Krass, Professor, Rotman School of Management, University of Toronto, 105 St. George Street, Toronto, ON, M5S 3E6, Canada, [email protected], Robert Aboolian, Oded Berman
We analyze the integrated location/pricing problem for a set of retail facilities on a network. All new facilities are assumed to charge the same price. Pre-existing competitive facilities may be present. Both the number and locations for the new facilities must be determined. We show that optimal pricing and location decisions can be separated, and pricing discretized, leading to efficient solutions.

I SD39
OK, Now How Do We Get There from Here?
Cluster: Where Do We Want to Go?
(In observance of Arthur Geoffrion’s retirement)

Invited Session
Chair: Kevin McCardle, Professor, UCLA Anderson School, 110 Westwood Plaza, Los Angeles, CA, 90095-1481, United States, [email protected] Co-Chair: Kumar Rajaram, UCLA Anderson School, 110 Westwood Plaza, Los Angeles, CA, 90095, United States, [email protected] 1 - Opportunities and Challenges for OR in the Next Five Years Karla Hoffman, Professor, George Mason University, Mail Stop 4A6, SEOR Department, 4400 University Drive, Fairfax, VA, 22030, United States, [email protected]
Earlier, OR was hampered by computing costs, lack of data and the inability of our tools to provide timely answers. Now, we are inundated with data, computing is virtually free, our tools are capable of providing real-time solutions, and graphical interfaces make it easy to control the process and output the findings. Together, these developments point to many new opportunities for OR. This talk highlights some new directions and challenges for OR modelers and educators.

4 - Constrained Center Problem with Unreliable Facilities Mozart Menezes, Assistant Professor, Department of Operations Management & Information Technology, HEC School of Management, Paris, 1 Rue de la Liberation, Jouy-en-Josas, 78351, France, [email protected], Oded Berman, Dmitry Krass
This paper is an extension of the p-Center problem on a network. A facility may be unavailable, causing customers to seek service from operating facilities. Thus, customers may have to travel to a facility that is not the closest one. Every customer requires that a facility will be available with some positive probability. A heuristic approach is developed to ensure coverage feasibility while minimizing the expected maximum distance traveled in order to obtain service.

117

SD41 I SD41
Learning and Adaptive Algorithms
Sponsor: Applied Probability Sponsored Session

INFORMS PITTSBURGH — 2006
This talk considers online stochastic combinatorial optimization (OSCO) problems where online decisions must select which requests to serve and how. These problems arise in many practical applications in networking, reservation systems, and vehicle routing and dispatching. This talk presents a class of anticipatory algorithms for OSCO applications, studies its theoretical properties, and demonstrates their applicability on a variety of complex problems.

Chair: Assaf Zeevi, Associate Professor, Columbia Business School, 3022 Broadway, New York, NY, 10027, United States, [email protected] 1 - A Universal Scheme for Learning Vivek Farias, Graduate Student, Electrical Engineering, Stanford University, 274 Packard Electrical Engineering, Stanford, CA, 94305, United States, [email protected], Ciamac Moallemi, Benjamin Van Roy, Tsachy Weissman
We consider an agent interacting with an unmodeled environment. At each time, the agent makes an observation, takes an action, and incurs a cost. Actions can influence future observations and costs. The goal is to minimize long-run average cost. We propose an algorithm inspired by the Lempel-Ziv scheme for data compression. We establish that if the future is conditionally independent of the past given a finite window of actions and observations, average cost converges to the optimum.

3 - Allocation of Jobs and Resources to Pooling Centers Hui-Chih Hung, Post-Doctoral Researcher, Ohio State University, 1971 Neil Avenue, Department of Industrial Engineering, Columbus, OH, 43210, United States, [email protected], Marc Posner
We partition identical and non-identical servers into parallel pooling centers and simultaneously assign job types to centers. Each job type has a distinct Poisson arrival rate and WIP weight. The goal is to minimize the total WIP cost. Heuristics are developed and are evaluated both theoretically and experimentally.

I SD43
AHP/ANP Theory and Applications
Cluster: Analytic Hierarchy Process Invited Session
Chair: Birsen Karpak, Professor of Management, Youngstown State University, One University Plaza, WCBA 635, Youngstown, OH, 44555, United States, [email protected] 1 - Success Factors for Small Medium Enterprises: An Analytical Network Process (ANP) Approach Birsen Karpak, Professor of Management, Youngstown State University, One University Plaza, WCBA 635, Youngstown, OH, 44555, United States, [email protected], Ilker Topcu
Small and Medium-sized Industries play important role in most economies. There are some studies done identifying the success factors of these organizations yet no study has been done yet prioritizing these factors. Considering interdependence of these factors, the authors developed Analytical Network Process (ANP) model which attempts to find importance of a variety of factors using expert judgments for a manufacturing industry.

2 - Using Proximity to Speed Up Experts Algorithms in Reactive Environments Daniela Pucci de Farias, [email protected], Jeremy Schwartz
”Experts algorithms” use a fixed set of strategies to make decisions sequentially in unknown environments. In reactive environments, the method matches the performance of the best expert, but bounds on convergence rate suggest a severe dependence on number of experts. We explore proximity metrics among experts to speed up convergence. A simple scheme is shown to asymptotically perform as well as the best expert. Empirical results suggest a much less dramatic dependence on number of experts.

3 - Risk Bounds and Near-Optimal Algorithms in Revenue Management with Unknown Demand Function Omar Besbes, PhD Student, Columbia Business School, 3022 Broadway, New York, NY, 10027, United States, [email protected], Assaf Zeevi
We consider a single product revenue management problem where realized demand is observed over time, but the demand function that governs these observations is not known. We define different information levels that characterize the a priori knowledge of the decision maker, and provide lower and upper bounds on the optimal performance of joint learning and pricing schemes. These results illustrate the risk-reward tradeoff with regard to the assumptions made on the underlying demand model.

2 - Analytic Network Process Model for Outsourcing Decisions Ozlem Arisoy, PhD Candidate, University of Pittsburgh, 1048 Benedum Hall, University of Pittsburgh, Pittsburgh, PA, 15232, United States, [email protected], Bopaya Bidanda, Larry Shuman
Outsourcing decisions encompass many different facets that are often assessed by stakeholders with conflicting objectives. In this study, we develop an Analytic Network Process (ANP) model that evaluates the multiple dimensions of outsourcing decisions within a corporate environment. The implementation of ANP is illustrated on a real life application and the results are evaluated by examining the fundamental principles of the methodology and elements of the model.

4 - Bandit Problems with Side Information Assaf Zeevi, Associate Professor, Columbia Business School, 3022 Broadway, New York, NY, 10027, United States, [email protected]
We consider a two armed bandit problem that serves as a canonical model of sequential decision making under uncertainty, and as such exhibits a clear trade off between estimation and optimization. We show that if one endows the decision maker with suitable ``side information,’’ then the optimal strategy and associated performance can be strikingly different than those seen in the classical setting where no side information is available.

3 - An Exploratory Study: Comparing Two Decision Making Methodologies Oyku Alanbay Isik, PhD Student, University of North Texas, College of Business Administration, Information Technology & Decision Sciences, Denton, TX, 76201, United States, [email protected]
An exploratory study was conducted to compare two software pieces, Expert Choice and Super Decisions, which are based on two different decision-making methodologies, Analytic Hierarchy Process(AHP) and Analytic Network Process(ANP).Applications were used with the same decision making problem. The findings indicate that software usability scale showed significant differences and respondents’ results with the applications were inconsistent. Various implications and future directions are suggested.

I SD42
Scheduling: New Models and Metrics
Cluster: Scheduling Invited Session
Chair: Jay Sethuraman, Associate Professor, Columbia University, 338 Mudd, IEOR Department, 500 West 120th Street, New York, NY, 10027, United States, [email protected] 1 - Characterizing the Effect of Inexact Size Information in Size-Based Policies Adam Wierman, [email protected]
When studying policies that “prioritize small jobs” theoriticians traditionally assume that policies have exact knowledge of job sizes, which is not true in practice. In this work, we define a class of policies that formalizes the heuristic of “prioritizing small jobs” in a loose enough manner so as to include policies that schedule using inexact job sizes. We then characterize the impact of inexact job size information by bounding the behavior of this classificiation.

I SD44
JFIG Paper Competition Finalists II
Sponsor: Junior Faculty Interest Group (JFIG) Sponsored Session
Chair: Alan Scheller-Wolf, Associate Professor, Tepper School of Business, Carnegie Mellon University, Pittsburgh, PA, 15213, United States, [email protected] 1 - JFIG Paper Competition Finalists
In these two sessions the finalists of the 2006 JFIG paper competition will present their research. The paper competition is organized by the Junior Faculty Interest Group (JFIG) forum in INFORMS. JFIG was created in 2001 to promote the career development of tenure-track faculty in INFORMS. The goals of the paper competition are to encourage research among junior faculty and to

2 - Anticipatory Algorithms for Online Stochastic Combinatorial Optimization Pascal Van Hentenryck, Professor of Computer Science, Brown University, Box 1910, 115 Waterman Street, 4th Floor, Providence, RI, 02912, United States, [email protected]

118

INFORMS PITTSBURGH— 2006
increase the visibility of research conducted by junior faculty within the fields of operations research and management science. Six finalists are selected by the award committee and their papers are presented in these two sessions. The papers submitted for the competition are evaluated based on the importance of the topic, appropriateness of the research approach, and the significance of research contribution. Winners are announced at the JFIG lunch on Sunday at the INFORMS Annual Meeting.

SD48

3 - Dynamic Allocation of Flexible Resources to Market Differentiated Stochastic Demand Eylem Tekin, Assistant Professor, Texas A&M University, Industrial Engineering, College Station, TX, 77843, United States, [email protected], Tao Huang
We consider a network of multiple flexible resources with fixed initial capacities that are used to satisfy demands from different market segments during a finite planning horizon. Using a dynamic programming approach, we show the structural properties of the optimal allocation policy for a two-resource case. For more general cases, we propose heuristic approaches derived from these structural properties.

I SD45
Tutorial: Electricity Deregulation: What is Wrong & How We Got There
Cluster: Tutorials Invited Session
1 - Electricity Deregulation: What is Wrong & How We Got There Lester Lave, Professor, Tepper School of Business, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected], Seth Blumsack
Regulated monopolies for electricity supply worked well from 1910 until 1975, when a host of problems led to deregulation. But deregulation has produced higher prices, unreliability, and transmission congestion. Recently the emphasis on deregulation has eased and some states have returned to their original regulatory regimes. We explore reasons for current problems in electricity markets and evaluate solutions. We conclude that either regime may satisfy market demands, though neither will produce a free competitive market.

4 - The Impact of Supply Quality Information and Supplier Development on Contract Design Sila Cetinkaya, Associate Professor, Texas A&M University, Industrial and Systems Engineering, College Station, TX, 77843, United States, [email protected], Xingchu Liu
We develop analytical models for designing optimal buyer-initiated supply contracts with supply quality and supplier development considerations while modeling private information and individual incentives explicitly. We use optimal control theory for computing the optimal contract parameters and provide a DPbased decision framework for managing supply quality information uncertainty.

I SD48
Game Theory Applications in Supply Chain Management
Sponsor: Manufacturing & Service Operations Management Sponsored Session
Chair: Ravi Anupindi, Associate Professor of Operations Management, Ross School of Business, University of Michigan, 701 Tappan Avenue, Ann Arbor, MI, 48109, United States, [email protected] 1 - Information Asymmetry and Advance Purchase Discounts in Supply Chains Krishnan Anand, Assistant Professor, Operations and Information Management, The Wharton School, 3730 Walnut Street, Suite 500, Philadelphia, PA, 19104, United States, [email protected], Karan Girotra
Advance Purchase Discounts (APD) have been shown to be beneficial in acquiring early demand information from customers ((Tang et al (2004)) and in coordinating two-tier supply chains (Cachon (2001)). In this study, we investigate the administration and benefits of APDs in two-tier supply chains where the two tiers have asymmetric information. We study the effect of APD contracts on the profits of the individual agents in the supply chain, as well as on total supply chain profits and welfare.

I SD46
Tutorial: Opt Art
Cluster: Tutorials Invited Session
1 - Opt Art Robert Bosch, Donald R. Longman Professor of Mathematics, Department of Mathematics, Oberlin College, Oberlin, OH, 44074, United States, [email protected]
Optimization is concerned with finding the best way to complete a task. Many applications of optimization, such as scheduling, matching and routing, are wellknown. In this tutorial, we will showcase novel applications of optimization to the area of fine art: portraits constructed out of complete sets of dominoes (via integer programming), mosaics comprised of abstract geometric tiles (via integer programming and various heuristics), and continuous line drawings (via the ``solution’’ of large-scale instances of the traveling salesman problem).

I SD47
Managing Uncertainty and Risk in Supply Chains
Cluster: Supply Chain and Operations Engineering Invited Session
Chair: Sila Cetinkaya, Associate Professor, Texas A&M University, Industrial and Systems Engineering, College Station, TX, 77843, United States, [email protected] 1 - Capacity Disruption Risk in Production Enterprise Networks Martin Wortman, Professor, Texas A&M University, Industrial and Systems Engineering, College Station, TX, 77843-3131, United States, [email protected], Alex Savachkin
The introduction of lean business practices can greatly enhance the operating efficiency of production enterprises; however, these practices can also leave production networks brittle with dramatically increased exposure to disruptions in both production and supply capacities. In this talk, we explore capacity disruption risk in large production networks. In particular, we examine structural properties of the underlying stochastic processes that characterize disruption risk.

2 - Inventory Assortment and Substitution Problems Yehuda Bassok, Professor, University of Southern California, Marshall School of Business, Los Angeles, CA, 90089, United States, [email protected], Feng Chen
We study a general substitution problem, in which consumers choose one of N variants. We start with a choice model that ranks the preference of each consumer. The preferences of the consumers are not known to the retailer and thus, he assumes that the demand for each variant is random. We derive the retailer’s optimal stocking policy. We show that the role of safety stock is to hedge against the uncertainty in the market size but not the uncertainty in the demand for each of the variants.

3 - Referral Infomediaries Under Demand Uncertainty and Finite Stock Li Jiang, Doctoral Student, Ross School of Business, University of Michigan, D0263 Davidson Hall, 1234, Ann Arbor, MI, 48109, United States, [email protected], Ravi Anupindi
We analyze the competitive dynamics of pricing and inventory strategies of two retailers who face demand from two customer segments - loyal and comparison shoppers. The two retailers reach the comparison shoppers by enrolling in a service with a “referral infomediary”. We derive the equilibrium market structure - in terms of retailers’ enrolment strategy with the infomediary - under various contracts.

2 - Lateral Transshipment Options in a 2-Stage Supply Chain Gary Gaukler, Assistant Professor, Texas A&M University, Industrial and Systems Engineering, College Station, TX, 77843, United States, [email protected]
In this talk, we model the behavior of a 2-stage supply chain in which lateral transshipments at the second stage installations are possible. Lateral transshipments can be used to quickly react to demand (and supply) uncertainty. We derive a compound inventory control policy to determine when transshipments should be initiated and evaluate the impact on system performance.

4 - Coalition Stability in Assembly Models Greys Sosic, Assistant Professor, University of Southern California, Marshall School of Business, Bridge Hall 401, Los Angeles, CA, 90089, United States, [email protected], Mahesh Nagarajan
In this talk, we examine a decentralized assembly system where n component manufacturers sell to a single downstream assembler who faces deterministic price sensitive demand. We look at three types of competition — Supplier Stackelberg, Vertical Nash, and Assembler Stackelberg — and in each type allow the suppliers to freely form coalitions among themselves. Using dynamic concepts of stability, we predict the structure of the stable supplier coalitions.

119

SD49 I SD49

INFORMS PITTSBURGH — 2006
2 - Locating Access Points to Support Voice over Wireless Local Area Frederick Kaefer, Associate Professor, Loyola University Chicago, 1 E. Pearson, Chicago, IL, 60611, United States, [email protected]
Using wireless local area networks to carry voice communications can save organizations money and support business applications. The problem of where to locate access points in a wireless local area network to support voice communications is studied. A model is developed to determine the number of access points and their locations given the number of data and voice users within a specified area.

History of OR and Management Science
Cluster: History of OR and Management Science Invited Session
Chair: Ernest Koenigsberg, Professor Emeritus, W. A. Haas School of Business, University of California at Berkeley, Berkeley, CA, 94720, United States, [email protected] 1 - Pat Rivett (1923-2005); British OR Pioneer Graham K. Rand, Senior Lecturer, Department of Management Science, Lancaster University Management School, Lancaster, LA1 4YX, United Kingdom, [email protected]
Pat Rivett was first secretary of the OR Society, built up the OR group at the National Coal Board, and founded the first Department of OR in the UK, at Lancaster. He played a leading role in the first International Conference in OR, leading to the creation of IFORS. Using his outstanding communication skills he persuaded private and public sector bodies to apply OR. The nature of OR in Britain, an emphasis on application and a society with a majority of practitioner members, is his legacy.

3 - Using Kriging to Estimate Network Performance in Wireless Local Area Networks Abdullah Konak, Assistant Professor, Penn State - Berks, Tulpehocken Road, Reading, PA, 19610, United States, [email protected]
This research proposes an approach based on Kriging, which is a regression technique used in geostatistics to interpolate data, to estimate the performance indicators of wireless local area networks such as signal strength, interference, and data rate. The approach aims to remedy some drawbacks of the existing site survey software.

2 - The History of OR as a Study of Method William Thomas, Graduate Student, Harvard University, 1 Oxford Street, History of Science Department, Cambridge, MA, 02138, United States, [email protected]
Some OR practitioners are fond of referring to the early wartime history of the discipline: OR as Blackett and Morse did it. Assuming the field has progressed, why are these roots still considered important? Drawing on early military studies and work done at MIT and Arthur D. Little, I emphasize the importance then attached to the freeform study of extant operational method as a hallmark of creative and vital work, rather than the application of some peculiar OR methodology.

4 - Designing Resilient Telecommunication Networks Michael Bartolacci, Associate Professor of IST, Penn State University - Berks, PO Box 7009, Reading, PA, 19610, United States, [email protected], Abdullah Konak
The authors propose an approach for designing resilient telecommunication networks based on genetic algorithms. A methodology for estimating traffic efficiency is also proposed.

I SD51
Implementation Issues in Data Mining/Machine Learning
Contributed Session
Chair: Onur Seref, PhD Candidate, University of Florida, 303 Weil Hall, Gainesville, FL, 32611, United States, [email protected] 1 - A Comparative Study of the K-Means Algorithm and the Normal Mixture Model for Clustering Dingxi Qiu, PhD Candidate, Northwestern University, 2145 Sheridan Road, Room C-217, Evanston, IL, 60208, United States, [email protected], Bruce Ankenman, Ajit Tamhane
The K-means algorithm and the normal mixture model estimated using the EM algorithm are two popular methods for clustering, but their performances have not been compared analytically. We first provide asymptotic analytical characterizations of the classification rules used by the two methods for the univariate and the bivariate cases. Next we use these characterizations to compare their expected misclassification rates. Simulation results are provided to verify the analytical results.

3 - Constructing an American Technology Policy: The Adoption of Operations Research in the United States, 1942-1952 Erik P. Rau, Instructor of History, Department of History and Politics, Drexel University, MacAlister Hall 3025, Philadelphia, PA, 19104, [email protected]
This paper argues that the arrival in the United States in 1942 marks a broader attempt by civilian scientists, military officers, and other professionals to institute a coherent technology policy during World War II. Victory in war vindicated the general consensus on government-industry-university coordination (despite deep disagreements over its implementation), thereby paving the road for OR’s postwar career as a tool in science and technology policy making.

4 - Elwood Buffa’s Pioneering Contributions to the Evolution of Production and Operations Management Martin Starr, Professor Emeritus, Crummer Graduate School of Business, Rollins College, 1000 Holt Avenue, Winter Park, FL, 32789, United States, [email protected], Jaya Singhal, Kalyan Singhal
Elwood S. Buffa, a pioneer in production and operations management (P/OM), passed away in the summer of 2005, leaving a valuable legacy. During his academic career Buffa made lasting contributions to teaching and research of P/OM. His 1961 textbook, Modern Production Management provided structure and foundation for many P/OM courses in business schools.

2 - A k-Means Clustering Procedure for Binary Classification Stephen France, PhD Candidate, RUTGERS Business School, PhD In Management Program, 180 University Avenue, Newark, NJ, 07102-1803, United States, [email protected], J. Douglas Carroll
We develop a procedure based upon k-means clustering to carry out binary prediction on large datasets. We account for different missing data patterns by using the EM algorithm and multiple imputation. We use multidimensional scaling (MDS) with the Hamming distance metric to transform nominal and ordinal data.

I SD50
Wireless Systems
Sponsor: Telecommunications Sponsored Session
Chair: Mahdi Nasereddin, Assistant Professor, Penn State University, Berks, Tulpehocken Road, PO Box 7009, Reading, PA, 19605, United States, [email protected] 1 - A Neural Network-Based Approach for Predicting Connectivity in Wireless Networks Mahdi Nasereddin, Assistant Professor, Penn State University, Berks, Tulpehocken Road, PO Box 7009, Reading, PA, 19605, United States, [email protected], Abdullah Konak, Michael Bartolacci
This research proposes a Connectivity Decision Support System based on connectivity maps generated by a neural network approach. The proposed approach creates a coverage map based on the signal strengths from active wireless users. These data are used to train a neural network to predict the signal strengths or coverage for locations for which no active user is reporting. In other words, a neural network fills in gaps in a coverage map for a given network connection point.

3 - Multivariate Regression Models with PE Random Errors and Subset Selection Using GA and ICOMP Minhui Liu, PhD Candidate, Department of Statistics & Management Science, University of Tennessee, 329 Stokely Management Center, Knoxville, TN, 37996, United States, [email protected], Hamparsum Bozdogan
We develop two novel multivariate regression models with Power Exponential (PE) random errors. Our first model assumes that the observations are independent and the second model assumes that the observations are dependent. We develop method of moments and the maximum likelihood methods to estimate the model parameters. The model selection criteria such as AIC and ICOMP for both models are derived. A genetic algorithm approach is used to obtain the estimates of the model parameters.

120

INFORMS PITTSBURGH— 2006
4 - A Bayesian Methodology for Semi-Automated Task Analysis Shu-chiang Lin, Instructor, Purdue University, School of Industrial Engineering, Grisso, 315 N Grant Street, West Lafayette, IN, 47907-2023, United States, [email protected], Mark Lehto
This 4 year study presents an effort-intensive field-based approach for the call center’s naturalistic decision making’s environment. A new idea by combining the Bayesian approach with task analysis methodology was proposed. Verbal information conversation narratives containing 165,000 words pertaining 55 printer models and 70 software and hardware issues were classified. The preliminary results suggested that a fuzzy Bayesian based tool would be able to learn and predict subtask categories.

SD53

4 - Pricing in a Duopoly with a Lead Time Advantage Victor Martinez de Albeniz, Assistant Professor, IESE Business School, Av. Pearson 21, Barcelona, 08034, Spain, [email protected]
We analyze the price competition between two suppliers offering two different lead-times to a buyer. The buyer replenishes inventory in an infinite horizon context, using a base-stock policy with each one of the suppliers. Based on the buyer’s behavior, the suppliers set static prices that maximize their long-term average profit. We study the equilibrium of the pricing game.

5 - Selective Support Vector Machines Onur Seref, PhD Candidate, University of Florida, 303 Weil Hall, Gainesville, FL, 32611, United States, [email protected], Oleg Prokopyev, Panos Pardalos, O. Erhun Kundakcioglu
In this study, we introduce a classification problem in which sets x(i) of k data points, x(i) = {x(i,1), ..., x(i, k)}, are given for each class. We introduce various extensions to the core quadratic programming problem of support vector machine classifiers, such that an optimal subset of t / k points is selected from each set x(i). We show our results on neural datasets in which we use our formulation to align individual single-trial time series with each other.

I SD53
Topics in Revenue Management and Pricing
Sponsor: Revenue Management & Pricing Sponsored Session
Chair: Christopher Anderson, Richard Ivey School of Business, University of Western Ontario, London, ON, Canada, [email protected] 1 - A Differential Duopoly Game with Path Dependent Demand Leo MacDonald, Richard Ivey School of Business, University of Western Ontario, London, ON, Canada, [email protected], Christopher Anderson, Henning Rasmussen
We develop a path dependent demand model of a competitive duopoly, with the problem formulated as a differential game. One area of research where path dependent demand is relevant is strategic consumer behaviour, a subject gaining prominence in the RM literature, particularly where e-commerce is a component. We set-up and discuss the solutions for the game as well as conditions for existence, and describe policy (pricing/inventory) implications and managerial insights.

I SD52
Competition and Pricing
Sponsor: Revenue Management & Pricing Sponsored Session
Chair: Serhan Duran, Graduate Research Assistant, School of Industrial & Systems Engineering, Georgia Institute of Technology, Atlanta, GA, 30332, United States, [email protected] Chair: Julie Swann, Assistant Professor, Georgia Institute of Technology, Industrial and Systems Engineering, Atlanta, GA, 30332, United States, [email protected] 1 - Optimal Supplier Contracts under Asymmetric Inventory Information Hao Zhang, Assistant Professor, University of Southern California, 3670 Trousdale Parkway, Bridge Hall 401G, Los Angeles, CA, 90089, United States, [email protected], Mahesh Nagarajan, Greys Sosic
We study a model in which a single supplier sells to a single retailer who faces stochastic demand and follows periodic review. We assume that the initial inventory and subsequent demand realizations are known to the retailer but unobservable by the supplier. Adopting the principal agent paradigm, we model the supplier’s problem as a constrained optmization problem and characterize its optimal contract. Further, we compare the performances of the optimal contracts with some simpler contracts.

2 - Price Variation and Release Strategies of Online Auctions Fredrik Odegaard, Sauder School of Business, University of British Columbia, 2053 Main Mall, Vancouver, BC, V6T1Z2, Canada, [email protected], Martin L. Puterman
We consider a seller who, using online auctions, wishes to sell a finite number items. The decision maker must decide when to start each auction. This decision involves a trade off between excess holding cost and price cannabilzation through simultaneous auctions. We formulate the problem as a MDP and derive some structural properties of the optimal release policy. We also include an empirical validation of model assumptions using data from several thousand online consumer electronics auctions.

2 - The Effect of Competition on the Perfomances of Customer Rebates and Retailer Incentives Ozgun Caliskan Demirag, Georgia Institute of Technology, Ferst Drive NW, Groseclose Room 347, Atlanta, GA, 30332, United States, [email protected], Julie Swann, Pinar Keskinocak
In some industries, such as automotive, production costs are largely fixed and manufacturers offer promotions to increase sales and revenues. We analyze the effect of competition on the promotional and ordering decisions in a two-stage supply chain using game theory to model the interactions between competing retailers and manufacturers. We focus on “retailer incentive” and “customer rebate” promotions and find the equilibrium decisions under different demand models and market conditions.

3 - Dynamic Pricing for Nonperishable Products: A Game Theoretic Approach Soheil Sidbari, Assistant Professor, Charlton College of Business, University of Massachusetts, 285 Old Westport Road, North Dartmouth, MA, 02747, United States, [email protected], David Pyke
In this study, we contribute to the dynamic pricing literature by developing a finite-horizon model for two firms offering substitutable and nonperishable products with different quality levels. Unlike the traditional studies in dynamic pricing, we assume that the products are nonperishable and can be stored by customers to be used in future. The stockpile of the products generated by customers affects the demand in future periods.

4 - Revenue Management with Last Minute Getaways Srinivas Krishnamoorthy, Assistant Professor, Ivey School of Business, The University of Western Ontario, 1151 Richmond Street North, London, ON, N6A 3K7, Canada, [email protected]
In recent years the travel industry has seen the development of a whole new market: the last minute leisure travelers. We show that the opportunity to sell “distressed” inventory through online channels to such last minute impulse travelers allows providers to increase revenues and load factor. While there is a chance of some revenue leakage because of travelers postponing purchases in anticipation of last minute sales, providers can overall benefit by this selling strategy.

3 - Dual Sales Channel Management with Availability-Based Kay-Yut Chen, Principal Scientist, Hewlett-Packard Laboratories, MS 1U-2, 1501 Page Mill Road, Palo Alto, CA, 94304, United States, [email protected], Murat Kaya, Ozalp Ozer
We present a game-theoretic model of manufacturer-retailer strategic behavior in which the manufacturer employs both a direct online channel and a retail channel. While making the purchase decision, consumers consider the waiting time in the direct channel, product availability in the retail channel. We identify three dual channel strategies for the manufacturer depending on product and market characteristics. We support our results with economics experiments with human subjects.

121

SD54 I SD54

INFORMS PITTSBURGH — 2006 I SD55
Computer Science - Applications to OR
Contributed Session
Chair: Eli Faulkner, Software Scientist/Mathematician, Quantum Leap Innovations, 3 Innovation Way, Suite 100, Newark, DE, 19711, United States, [email protected] 1 - The Non-Stationary Data-Driven Newsvendor Problem Gokhan Metan, PhD Student, Lehigh University, 200 W. Packer Avenue, Mohler Lab, Bethlehem, PA, 18015, United States, [email protected], Aurelie Thiele
We investigate the impact of the sample size in the non-stationary newsvendor problem when the underlying demand distribution is not known, and the performance is measured by the decision-maker’s average regret. We propose an iterative algorithm to determine the number of past observations that should be included in the decision-making process, provide insights into the optimal sample size and perform extensive computational experiments.

Interfaces Between Marketing, Operations and Technology II
Cluster: Operations and Marketing for Emerging Markets Invited Session
Chair: Jiong Sun, Doctoral Student, Tepper School of Business, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected] Co-Chair: Sunder Kekre, Professor of Operations Management and Manufacturing, Tepper School of Business, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected] 1 - The Impact of Manufacturing Offshore on Technology Trajectories Erica Fuchs, Post-Doctoral Fellow, Massachusetts Institute of Technology, Building E38-104, 77 Massachusetts Avenue, Cambridge, MA, 02139, United States, [email protected]
This paper presents a two-case study of the impact of manufacturing offshore on the technology trajectory of the firm and the industry. It looks in particular at the automotive and optoelectronics industries. In both cases, manufacturing offshore changes which design is most economic. However, only in the optoelectronics case do the results support that manufacturing offshore is changing the path of technology development. The paper proposes a general framework to explain this difference.

2 - Real-Time Business Process Management System Dong-Ho Kim, u-Logistics Research Team, ETRI, 161 Gajeong-Dong, Yuseong-Gu, Daejeon, 305700, South Korea, [email protected], Sung Whan Choi, Sung Wook Kim
RBPMS receives a continuous stream of filtered data from RFID Middleware and produces events that trigger business processes based on a set of conditions. Our system provides a simple, yet powerful management support for a creation and monitoring of rule objects, hence enables business experts to flexibly adapt to rapidly changing business environment and automates business process as well as data flow when adapted to RFID-based system.

2 - Design of Extended Warranties in Supply Chains Suman Mallik, Assistant Professor, University of Illinois at UrbanaChampaign, 1206 South Sixth Street, 350 Wohlers Hall, Champaign, IL, 61820, United States, [email protected], Dilip Chhajed, Kunpeng Li
Consider a supply chain involving a manufacturer and a retailer. The manufacturer produces a single product and sells it through the retailer. The extended warranty could, however, be offered by either party. We use game theoretic models to answer the following questions. What are the characteristics of optimal extended warranty policy decisions? Which scenario leads to a higher total supply chain profit, a retailer offering the extended warranty or the manufacturer?

3 - A RFID-Based Postal Unit Load Device Management System Dong-Ho Kim, u-Logistics Research Team, ETRI, 161 Gajeong-Dong, Yuseong-Gu, Daejeon, 305700, South Korea, [email protected], Hong-Suk Hu, Sung-Woo Jun
The Korean postal e-logistics information system(PostNet) processes very large volume of information under the bar-coding mechanism environment. The research activities are still required in order to enhance the degree of postal automation using the ubiquitous information technologies. In this paper, we propose a new postal information system architecture based on RFID technology in postal unit load device so that enhance business efficiency and overall performance.

3 - Balancing Marketability and Manufacturability in Product Line Design Optimization Jeremy Michalek, Assistant Professor, Carnegie Mellon University, Scaife Hall 323, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, [email protected]
Product development involves communication and compromise among interacting and often competing objectives from marketing, design, and manufacturing perspectives; particularly when developing product lines for heterogeneous markets. This study demonstrates a methodology for explicitly quantifying tradeoffs between market performance and manufacturing cost for a line of consumer durables to achieve realizable product line solutions with optimal profitability.

4 - Internet Routing Under Active Congestion Control Stanko Dimitrov, PhD Student, University of Michigan, Industrial and Operations Engineering, 1205 Beal Avenue, Ann Arbor, MI, 48109, United States, [email protected], Dushyant Sharma, Marina Epelman
We present a mathematical model of Internet routing that incorporates congestion control techniques such as random early drop (RED). We show that the model is NP-hard and present numerical results for a continuous non-convex and a linear-integer formulation.

5 - The Adaptive Optimization Engine Eli Faulkner, Software Scientist/Mathematician, Quantum Leap Innovations, 3 Innovation Way, Suite 100, Newark, DE, 19711, United States, [email protected], Julia Cowart
The Adaptive Optimization Engine (AOE) solves nonlinear optimization and constraint satisfaction problems by interleaving 30 different algorithms ranging from heuristic search to generic algorithms to linear programming. Black box modeling and collaborative problem solving approaches allow modelers to represent their problem without worrying about the underlying solving technique. We present the AOE architecture and provide real world examples demonstrating its power and ease of use.

122

Sponsor Documents

Or use your account on DocShare.tips

Hide

Forgot your password?

Or register your new account on DocShare.tips

Hide

Lost your password? Please enter your email address. You will receive a link to create a new password.

Back to log-in

Close