Data Quality Based Data Integration Approach

 

World of Computer Science and Information Technology Journal (WCSIT)  ISSN: 2221-0741 Vol. 5, No. 10, 155-164, 2015

Data Quality Based Data Integration Approach

Mohamed Samir Abdel-Moneim

Ali Hamed El-Bastawissy

Mohamed Hamed Kholief

College of Computing and Information Technology Arab Academy for Science Technology & Maritime Transport Cairo, Egypt

Faculty of Computer Science MSA University Giza, Egypt

College of Computing and Information Technology Arab Academy for Science Technology & Maritime Transport Alexandria, Egypt

Abstract —   — D Data ata integration systems (DIS) are systems where query answers are collected from a set of heterogeneous and autonomous data sources. Data integration systems can improve results by detecting the quality of the data sources and retrieve answers from the significant ones only. The quality measures of the data in the data sources not only help in determining the significant data sources for a given query but also help data integration systems produce results in a reasonable amount of time and with less errors. In this  paper, we perform perform an experiment that shows shows a mechanism used to calculate and store a set of quality mea measures sures on data sour sources. ces. The quality measures are, then, interactively used in selecting the most significant candidates of data sources to answer user s’ s’  queries. The justification and evaluations are done using amalgam and THALIA benchmarks. We show that our approach dramatically improves query’s answers.  Keywords-component Keywords-com ponent data integration; quality measures; data sources; query answers; user preferenc preferences. es.

I. 

I NTRODUCTI  NTRODUCTION ON 

Data Integration (DI) is the process of combining the data located at multiple locations, and allowing the user to view these data through a single unified view called global or mediated schema [1, 2]. The global schema is the interface where users submit their queries to a data integration system. The user no longer needs to know how to access the data sources, nor does he need to consider how to combine the results from different sources. The data requested by the user may be found at a single source, at many sources, or scattered across many sources. Different architectures for data integration systems have  been proposed, proposed, but broadly speaking, most sy systems stems fall between warehousing and virtual integration [3]. The quality of the data sources can dramatically change as data may be incomplete, inaccurate or out of date. In fact, the quality of the result depends mainly on two factors: the quality of the data at the data sources and the manipulation process that  builds the resulting data from the data sources sources.. Because the quality of the data sources can dramatically change, it is important to store some quality-related measures about the data sources to take it into consideration during query planning.

data integration system. Attribute values can be integrated from different data sources based on quality measures and user ’s   preferences.  preferenc es. We use quality measures to deliver query answers with satisfied quality. In this paper we perform an experiment that is based on that work [4]. The experiments were conducted using two publicly available benchmarks for data integration systems: Amalgam Integration Test Suite [5] and Test Harness for the Assessment of Legacy information Integration Approaches (THALIA) [6]. The work performed is not a complete data integration system. Rather, it’s an extension to any data integration system. The rest of this paper is organized as follows. In Section II, we briefly discuss the data quality dimensions used in our  previous work. Section III illustrates the architecture and functions of our data integration quality system components. Section V describes our quality driven query processing algorithm. The experiments are described in section VI. The conclusion and future work are presented in Section VII. This work is part of a complete research group composed of researches from Cairo University and Arab Academy for Science Technology & Maritime Transport (AAST) focusing on data integration topics [4, 7, 8, 9, 10].

In our previous work [4], we presented an approach that is  based on utilizing data quality (DQ) aspects in data integration systems in order to get satisfied query plans. Our approach is  based on adding quality system components to be parts of any 155 

 

WCSIT 5 (10), 155 -164, 2015

II. 

 E.  Timeliness

DATA QUALITY DIMENSIONS USAGE IN DATA INTEGRATION  

Timeliness is how old the data are in a data source [14]. Timeliness is important as some data sources might be outdated and the user might be interested in getting up-to-date up -to-date data.

In general, data quality means “fitness for use” [11, 12]. So, the interpretation of the quality of data item depends on the user’s needs. Wang and Strong [13] have empirically defined fifteen data quality dimensions considered by end users as the most significant. They classify these dimensions into contextual, intrinsic, representational and accessibility quality as shown in “Figure 1”. 

III. 

QUALITY SYSTEM COMPONENTS  

The quality system component consists of (1) Data quality acquisition and (2) user input. The quality system components are integrated in the mediator-wrapper architecture. See green  boxes in “Figure 2.” 

Figure 1. A conceptual framework of data quality

In our previous work, we selected data quality dimensions that could affect the data integration process and could be considered important from user ’s ’s prospective. The data quality dimensions chosen were:  A.   Accuracy

Wang and Strong [13] defines accuracy as “The extent to which data are correct, reliable, and certified free of error”.   Increasing accuracy of the query answer is importa nt from user’s  prospective as data sources might contain incorrect or misspelling data.

Figure 2. Data integration system quality system components

In the following sub-sections, we present the structure and the functionality of each component.

 B.  Completeness

Completeness defined as “the extent to which data are of sufficient breadth, depth, and scope for the task at hand” [13]. Querying one data source gives a set of results. As the number of data sources queried increase, the result will be more complete. C.  Cost

Cost is the amount of money required for a query. Considering cost is important so that users can choose between free and commercial data sources.

 A.   Data Quality Acquisition The data quality acquisition (DQA) component is responsible for extracting attributes and relations from the t he data sources and store them in the metadata store. It is also responsible for executing data quality queries against the data sources, receiving the results and store them in the metadata store. The metadata store used by the DQA is shown in “Figure 3” :

 D.   Response Time

It is the amount of time when the mediator submit a query and receive the complete response from the data source. Response time is important as users waiting a long time for a response are more willing to abandon the query.

156

 

WCSIT 5 (10), 155 -164, 2015

various measurement techniques based on choices made regarding four factors: where to measure the data, which part of the data will be measured, how to measure the data and the granularity of the measures. We applied Redman’s data accuracy measurement framework in our case by selecting from the choices for each of the four factors.

  Where measurements are taken: We measured accuracy



from the data sources. (i.e. from database).

  What attributes to include: We measured accuracy on the



data sources’ attributes that correspond to global schema’s attributes. 

  The measurement device: We will compared the value of



Figure 3. Metadata structure

each attribute to its domain of allowed values. Complaints and domain experts’ feedback were also used to identify erred data and a correction for them which help improve accuracy measure.

Table I illustrates the data quality dimensions selected in our  previous work work and the granularity granularity for each dimension. TABLE  I.

DATA QUALITY DIMENSIONS AND GRANULARITIES LEVELS  

DQ Dimension

Accuracy   Completeness Cost Response time Timeliness

  The scale on which results are reported: Attribute level.



Measures granularities D ata source level

R elation level

A ttri bute level

Attribute Accuracy =

   

 Number of fields judget correctly  Number of fields tested

 

(1)

2)  Completeness The Literature classifies completeness into three types: column completeness, schema completeness, and population completeness [16]. At the most abstract level, schema completeness refers to the degree to which all required information are present in a particular data set. At the t he data level, column completeness can be defined as the measure of the missing values for a column in a table. Each of the three types can be measured by dividing the number of incomplete items by the total number of items and subtracting from 1 [16].

     

In the following sub-sections, we describe how we measure each dimension presented in table I. These quality measures may enhance the quality of the data fusion process. Data quality dimensions chosen are highlighted in blue in “Figure 3”. 1)   Accuracy Tomas C. Redman [15] present the data accuracy measurement framework (“Figure 4”) for understanding the

Schema/Attribute completeness = 1 −

 Number of incomplete items Total number number of items

 (2)

Figure 4. The data accuracy measurement framework 

The range for completeness is 0 - 1, where 0 represents the lowest score and 1 represents the high score. We add a custom data quality criteria cal led “Complete instance relation” that can be measured at schema level.  A relation is marked as complete instance if its cardinality is

complete. (i.e. all the tuples are represented in the relation). This information will be given directly to the data integration system  by the end user through through an inp input ut form.

157

 

WCSIT 5 (10), 155 -164, 2015 3)  Cost It is the price for accessing specific data source. We assume that the user is charged on pay-by-query basis. The cost per query is measured in US dollar. 4)   Response Time We measured the response time of a data source by sending a bunch of queries to the data sources to judge their average response time for different types of queries at different times of day.

query algorithm presented in [4] to determine which combinations of sources can answer the query. V. 

In this section, we describe the implementation of the query  planning algorithm and the quality system components. The goals of the experiments are to measure the response time, number of data sources needed to answer a given query and the cost of accessing the data sources to answer the queries. The experiments were done according to the following execution  paths:

5)  Timeliness We measured timeliness by using the update information  provided by the data source. We assumed that the data source updates its data at the relation level and the data at the data sources are not archived.

When no quality measure were calculated. (i.e. the data integration system ignores the pre-saved quality measures as if they weren’t exist) 

 

Default execution. This means if the user didn’t specify quality constraints, the DIS retrieves the best result according to the pre-saved quality measures.

 

When user specifies quality constraints. In this case, the user has selected some attributes and specified quality constraints on some of them.



To give users the option to specify constraints on the retrieved result, we used the proposal of Gertz and Schmitt [17]. We added two options to the SQL dialect. The first one is cost which is the amount of money a user can pay and the second option called fusion that can be set to true or false and is used to give the user the option to retrieve data from all possible data sources.



Also during the experiments, all attributes from the global schema were selected.

A query Q with quality constraint expressed on the mediated schema expressed in an extended SQL syntax:

We ran the experiments on a laptop shipped with an Intel Core i7-2760QM with 4 x 2.4 GHz CPU and 6 GB RAM. The laptop operates with Windows 7 ultimate edition. The tools used for the experiments were Microsoft SQL Server 2014® and Microsoft visual C# 4.5®.  A.   Amalgam

Amalgam is a benchmark which consists of several schemas and datasets storing bibliographic information. It consists of four schemas. Each schema represents a data source. The authors of the benchmark require anyone who needs the data to request it from them. So, we requested the data from the authors and gratefully received it. We created the schemas in a SQL server database called “Amalgam” and loaded the data into it.

Selection condition: conditions used to filter the data. Data quality goal: quality dimensions defined on the selected attribute Ai and gets a value according to table II. TABLE II.

DATA QUALITY DIMENSIONS LEVELS  

Level High

 Start thre hresho shold

Meduim Low

50 0

The first component of our quality system components is the t he data quality acquisition component. As we mentioned in section III, the data quality acquisition component is responsible for

70

extracting and relations sources and store themattributes in the metadata store. from It is the alsodata responsible for executing data quality queries against the data sources, receiving the results and store them in the metadata store. So, we created the metadata store described in “Fig 3”  which consists of six tables in the same database “Amalgam”.   We created a tool to map the global schema columns with the local schema columns. Table III shows the global schema tables and global schema columns used:

Fusion: When set to true, this means that the user wants to fuse data from all possible data sources. When set to false, the mediator selects only one alternative that has the minimum number of data sources. Cost: the amount in US dollar the user can pay. IV. 

 



 B.  User Input

Select A1,…..,Ak   from G where < selection condition > with < data quality goal > fusion < true | false > Cost < x$ > Where A1.A2,…, Ai are global attributes of G

EXPERIMENTS AND RESULTS 

QUALITY DRIVEN QUERY PROCESSING  

TABLE III.

The data requested by the user is usually located on more than one data source. Every combination of data sources that meet the user’s requirements (attributes and quality criteria) is an alternative. If a single data source can meet all user’s requirement, this is an alternative. Given a query Q against the mediated schema asking for A1,…..,An attributes with or without quality requirements, We developed a quality-driven

GLOBAL SCHEMA TABLES AND COLUMNS  

Global Schema Table

Article  Book

158

Global Schema Columns

ArticleID Month Title

Title Pages Publisher

Author Volume Year

Journal Location Month

Year Abstract Pages

 

WCSIT 5 (10), 155 -164, 2015

The data quality queries executed by the data quality acquisition component were implemented as stored procedures. Those stored procedures contain the equations used to calculate the completeness and accuracy of the attributes in the data sources. The stored procedures were ran according to a SQL server scheduled scheduled job. The job can be customized by tthe he system administrator according to the data sources change frequency. Also we can change the queries used by the data quality acquisition anytime. Whenever data quality acquisition completes a run, the quality measures in the metadata store will

The results in table V show that response time is reduced after adding the quality measures even if the user did specify quality criteria regardless of the fusion option.  b)   Number of accessed accessed data sources sources Table IV shows the number of accessed data sources needed to answer the query. TABLE VI. 

 be updated with the new values. The second component is the user input. The purpose of the user input component is to give the users the option to specify quality constraints on the retrieved result. The user selects the required attributes and optionally specify a data quality constraint on each selected attribute. The user can choose  between accuracy accuracy and complet completeness. eness. The user also also has to select the level of the DQ constraint which can be: high, medium or low. These levels get values according to table II. Also the t he user can check the fusion option and specify a cost of accessing a data source in case there are data sources that require a cost.

Execution path

 No quality measure User specified quality Default

Data Source S1

S2 S3 S4

THE QUALITY MEASURES OF THE DATA SOURCES  Complete instance tables Article, Author authors, citForm, journal, abstracts, months, numbers,  pages, titles, titles, vol volumes, umes, years years author, article author , publication

Cost 3$

Response time 500 sec

2$

500 sec

4$ 5$

500 sec 500 sec

TABLE VII. 

 No quality measure

a)  Response time

User specified quality

Table V shows the response time of our approach in both scenarios (when fusion is false and true) and according to the different execution paths:

 No quality measure measure User specified quality Default

Default

Title, Journal year -

Response time (sec)

Fusion = false

Fusion = true

1.445 sec

1.455 sec

0.772 sec

1.328 sec

0.458 sec

1.408 sec

Fusion = true

4

4

2

4

2

4

COST NEEDED TO ANSWER THE QUERY  Attributes with DQ constraints

Title, Journal year -

Cost($)

Fusion = false

Fusion = true

14

14

5

14

5

14

The cost results table VIItoshow thatassumptions if the data in sources do require andinaccording the cost table IV, the amount of money is reduced after adding the quality measures. However, it remains the same when fusion was set to true, because when fusion set to true, the query planning may access the whole data sources according to the alternatives generated.

R ESPONSE ESPONSE TIME 

Attributes with DQ Constraints

Fusion = false

Table VII shows the amount of money needed to answer the query.

The cost criteria selected were 7$. Hence, all data sources will be used to answer the query.

Execution path

Title, Journal year -

Number of accessed data sources

consisted of the 4 data sources. c)  Cost of answering the query

Execution path

TABLE V. 

Attributes with DQ constraints

The results in table VI show that if no quality measures were added, the DIS needs to query the whole data sources. While after adding quality measures, the number of data sources reduced to 2 instead of 4. The number of data sources remain 4 when fusion was set to true, because the query planning algorithm merged the data sources in all alternatives and queried each data source only once. The alternatives generated were

We considered two different scenarios w.r.t fusion option. In the first scenario, the fusion option is set to false while in the second is set to true. Regardless of the scenarios, table IV shows the quality measures of the data sources: TABLE IV. 

 NUMBER OF ACCESSED DATA SOURCES  

The following charts represent the results of amalgam  benchmark:

 

159

 

WCSIT 5 (10), 155 -164, 2015

Response Time Chart 1.6

1.455 1.445

1.408

1.328

1.4 1.2 1

0.772

0.8

Response Time Fusion = False 0.458

0.6

Response Time Fusion = True

0.4 0.2 0 No qual qualit ity y me meas asur ure e

Us User er req requi uire red d quality

Default

Number of Accessed Data Sources Chart 4.5

4

4

4

2

2

4 3.5 3 2.5

No. of Accessed Data Sources

2 1.5

Fusion = False No. of Accessed Data Sources Fusion = True

1 0.5 0 No qual qualit ity y me meas asur ure e

Us User er requ requir ired ed quality

Default

Cost Chart 16

14 14

14

14

14 12 10 Cost Fusion = False

8 6

5

5

U ser re required qu qua lity

Def a ul t

Cost Fusion = True

4 2 0 No qu qua lity m me e a s ure

 B.  THALIA

THALIA (Test Harness for the Assessment of Legacy information Integration Approaches) is a public available testbed and benchmark for information integration systems. It

goal of the benchmark is a systematic classification of the different types of syntactic and semantic heterogeneities that are described by the twelve queries provided.

 provides 42 downloadable university course catalog from computersources science representing around the world. The

As webydid in amalgam benchmark, all schemas and data  provided THALIA benchmark were loaded into a relational

160

 

WCSIT 5 (10), 155 -164, 2015 database. The database called “Thalia”.  For the data quality acquisition component, we created the metadata store described in “Fig 3” which consists of six tables in the same database “Thalia”. We used the same mapping tool we used in amalgam  benchmark to map the global schema columns with the local schema columns. Table VIII shows the global schema table and global schema columns used: TABLE VIII. 

GLOBAL SCHEMA TABLES AND COLUMNS  

Global Schema Table

Global Schema Columns

Code Prerequisite Prerequisi te HomePage

Course 

CourseName Days Description

Instructor Building

The second component is the user input. As with amalgam  benchmark, the user can choose between accuracy and completeness and the level of the DQ constraint which can be: high, medium or low. Also the user can check the fusion option and specify a cost of accessing a data source in case there are data sources that require a cost.

THE QUALITY MEEASURES OF THE DATA SOURCES  Is complete instance

Cost

Response time

Arizona State University

True

3

500

Bilkent University

False

3

500

Bosphorus University

True

3

500

Boston University

False

3

500

Brown University

False

3

500

False

3

500

False

3

500

Columbia University

False

3

500

Cornell University

False

3

500

True

3

500

True

3

500

Florida State University

True

3

500

Furman University

False

3

500

Georgia Tech

False

3

500

Harvard University

False

3

500

Hong Kong University

False

3

500

California Institute of Technology Carnegie Mellon University

Eidgenössische Technische Hochschule Zürich Florida International University

3

500

Kansas State University

False

3

500

Michigan State University

False

3

500

MiddleEast Technical University

False

3

500

NewYork University

False

3

500

Northwestern University

False

3

500

False

3

500

True

3

500

Stanford University

False

3

500

UniversidaddePuertoRico Bayamon

False

3

500

University of Arizona

False

3

500

University of Berkeley

False

3

500

True

3

500

True

3

500

University of Florida

False

3

500

University of Illinoisat UrbanaChampaign

False

3

500

University of Iowa

False

3

500

University of Maryland

False

3

500

University of MassachusettsBoston

False

3

500

University of Michigan

True

3

500

University of NewSouth Wales Sydney Australia

False

3

500

University of Toronto

False

3

500

University of Virginia

True

3

500

Washington University

True

3

500

Worcester Polytechnic Institute

False

3

500

Yale University

False

3

500

University of California Los Angeles University of California SanDiego

We considered two different scenarios w.r.t fusion option. In the first scenario, the fusion option is set to false while in the second is set to true. Regardless of the scenarios, table IX shows the quality measures of the data sources:

Data Source 

False

Pennsylvania State University Rochester Institute of Technology

Credits Room

The first component of our quality system components is the data quality acquisition component. As with amalgam  benchmark, the data quality queries queries execu executed ted by the data quality acquisition component were implemented as stored procedure procedures. s.

TABLE IX. 

Istanbul Technical University

Since each data source has one table, the complete instance table measure is attached to the data sources. The cost criteria selected were 7$. Hence, all data sources will be used to answer the query. a)  Response time Table X shows the response time of our approach when fusion is false and true and according to the different execution  paths: TABLE X. 

161

R ESPONSE ESPONSE TIME  

 

WCSIT 5 (10), 155 -164, 2015

Execution path

 No quality measure measure User specified quality Default

Attributes with DQ constraints

The results in table XI show that if no quality measures were added, the DIS needs to query the whole data sources. While after adding quality measures, the number of data sources reduced to 5 instead of 42 when fusion was false and to 11 when fusion was true.

Response time (sec)

Fusion = false

Fusion = true

CourseName, Code, Instructor

1.723 sec

1.716 sec

0.831 sec

1.801 sec

-

0.774 sec

1.706 sec

c)  Cost of answering the query Table XII shows the amount of money needed to answer the query. TABLE XII. COST NEEDED TO ANSWER THE QUERY  

The results in table X show that response time is reduced after adding the quality measures when fusion was false even if the user did specify quality criteria. When fusion option was true, it required a little time because the quality constraints checks. However, the default execution requires time less than when no quality measures were calculated.

Execution path

 No quality measure User specified quality

 b)   Number of accessed accessed data ssources ources

Default

Table XI shows the number of accessed data sources needed to answer the query: TABLE XI. 

Execution path

 No quality measure measure User specified quality Default

Number of accessed data sources Fusion = false

Fusion = true

42

42

5

11

5

11

Title, Journal year -

Title, Journal year -

Cost($)

Fusion = false

Fusion = true

126

126

15

33

15

33

The results in table XII show that if the data sources do require cost and according to the cost assumptions in table IX, the amount of money is reduced after adding the quality measures regardless if fusion option. When fusion is true, the query planning accesses all data sources in all alternatives generated. That’s why the cost is high when fusion is true.

 NUMBER OF ACCESSED DATA SOURCES   Attributes with DQ constraints

Attributes with DQ constraints

The following charts represent the results of THALIA  benchmark:

Response Time Chart 2

1.723 1.716

1.801

1.706

1.5 0.831

1

0.774

Response Time Fusion = False Response Time Fusion = True

0.5 0 No qual qualit ity ym mea easu sure re

U sser er requ requir ired ed quality

Default

 

162

 

WCSIT 5 (10), 155 -164, 2015

Number of Accessed Data Sources Chart 42

45 40 35 30

Number of Alternatives Fusion

25 20 15

11

11

10

5

5

U sser er requ requir ired ed quality

Default

= False Number of Alternatives Fusion = True

5 0 No qual qualit ity ym mea easu sure re

Cost Chart 140

126

120 100 80

Cost Fusion = False

60 40

33

33

15

15

U s er r eq u ir ed q u a l i ty

D ef a ul t

Cost Fusion = True

20 0 No qua l it y m ea s ur e

R EFERENCES EFERENCES  VI. 

CONCLUSION AND FUTURE WORK

Data integration systems may suffer from producing results that not only lack the quality but also take a long time to arrive. In this paper, we have pointed out the importance i mportance of data quality in integrating autonomous data sources. The main contribution of this paper is an efficient method aimed at selecting a few  possible data sources to provide more quality oriented result to the user. We added quality system components to integrate data quality dimensions in a data integration environment for structured data sources only. With the help of these criteria, we developed a quality driven query execution algorithm to generate high quality plan that meets user’s requirements.  Our experiments show that our approach delivers result in a reasonable amount of time and using the minimum number of data sources possible. Further research will ex extend tend the approach to be applied on different types of data sources such as semi-structured and unstructured data sources.

[1]

M. Lenzerini, Lenzerini, "Data integration: a theoretical perspective," in Proceedings of the twenty-first ACM SIGMOD-SIGACT-SIGART symposium on Principles of database system, Madison, Wisconsin, June 03-05, 2002.

[2]

A. Y. Halevy, "Answering queries using views: A survey," The VLDB Journal  —   The International Journal on Very Large Data Bases, vol. 10, no. 4, pp. 270-294, Decem December ber 2001.

[3]

A. Doan, A. Halevy and Z. Ives, Principles of Data Integration, San Francisco, CA: Morgan Kaufmann Publishers Inc., 2012.

[4]

M. Samir, A. H. El-Bastawissy and M. Kholeif, "Quality Driven Approach for Data Integration Systems (http://icit.zuj.edu.jo/icit15/DOI/Database_and_Data_Mining/0078.pd f)," in The 7th International Conference on Information Technology, Amman, Jordan, 2015.

[5]

R. J. Miller, D. Fisla, M. Huang, D. Kalmuk, F. Ku and V. L Lee, ee, "The Amalgam Schema and Data Integration Test Suite http://dblab.cs.toronto.edu/~miller/amalgam/," 2001.

[6]

J. Hammer and M. Stonebraker, "Thalia: Test harness for the assessment of legacy information integration approaches," in In Proceedings of the International Conference on Data Engineering (ICDE) Pages 485-486, 2005.

[7]

A. Al-Qadri, A. H. El-Bastawiss El-Bastawisssy sy and O. M. Hegazi, "Approximating Source Accuracy Using Dublicate Records in Data Integration," IOSR

163

 

WCSIT 5 (10), 155 -164, 2015

Journal of Computer Engineering (IOSR-JCE), vol. 13, no. 3, pp. 6872, 2013. [8]

M. M. Ha Hafez, fez, A A.. H. El-Bastawissy El-Bastawissy and O. M. Hegazy, Hegazy, "A stati statistical stical data fusion technique in virtal data integration environment," International Journal of Data Mining & Knowledge Management Process (IJDKP), vol. 3, no. 5, pp. 25-38, September 2013.

[9]

M. M. Hafez, A. H. El-Bastawissy and O. M. Hegazy, "Using Information Gain in Data Fusion and Ranking," in Recent Advances in Computer Engineering, Communications and Information Technology,  pp. 157 – 165, 165, 2014.

[10] A. Z. El Qutaany, A. H. El Bastawissy and O. Hegazy, "A Technique for Mutual Environment," InconsistenciesinDetection andand Resolution Virtual 2010 Data Integration Informatics Systems in (INFOS), The 7th International Conference on, 2010. [11] J. M. Juran, The The Quality Control Handbook, 3rd ed., New York: McGraw-Hill, 1974. [12] G. Kumar Tayi and D. P. Ballou, Examining data quality, Communications of the ACM,, v.41 n.2, p.54-57, Feb. 1998. [13] R. Y. Wang and D. M. Strong, "Beyond accuracy: what data quality means to data consumers," Journal of Management Information Systems, vol. 12, no. 4, pp. 5-33, March 1996. [14] F. Naumann, Quality-Driven Query Answering for Integrated Information Systems, Springer-Verlag Berlin, Heidelberg, 2002. [15] T. C. Redman, "Measuring Data Accuracy A Framework and Review," in Information Quality, R. Y. Wang, E. M. Pierce, S. E. Madnick and C. W. Fisher, Eds., Armonk, NY, M.E. Sharpe, 2005, pp. 21-36. [16] L. L. Pipino, Y. W. Lee Lee and R. Y. Wang, "Data quality assessment," Communications of the ACM, vol. 45, no. 4, pp. 211-218, April 2002. [17] M. Gertz and I. Schmitt, "Data integration techniques based on data quality aspects," in 3rd National Workshop on Federal Databases, Magdeburg, Germany, 1998. [18] G. Wiederhold, Wiederhold, "Mediators in the Architecture of Future Information Systems," IEEE Computer, vol. 25, no. 3, pp. 38-49, March 1992. [19] Y. Wand and R. Y. Wang, "Anchoring data quality dimensions in ontological foundations," Communications of the ACM, vol. 39, no. 11,  pp. 86-95, November November 199 1996. 6. [20]

M. Spiliopoulou, "A calibration mechanism identifying the optimization technique of a multidatabase participant," in Proc. of the Conf. on Parallel and Distributed Computing Systems (PDCS), Dijon, France, September 1996.

[21] M. Scannapieco, A. Virgillito, C. Marchetti, M. Mecella and R. Baldoni, "The daquincis architecture: a platform for exchanging and improving data quality in cooperative information systems," Information Systems, vol. 29, no. 7, pp. 551 - 582, October 2004. [22] F. Naumann, U. Leser Leser and J. C. Freyta, "Quality-driven integration of heterogenous information systems," in 25th proceeding of the International Conference on Very Large Databases (VLDB) , p.447458, Edinburgh, Scotland, September 07-10, 1999. [23] M. Ge and M. Helfert, "A Review of Information Quality Research Develop a Research Agenda," in Proceedings of the 12th International Conference on Information Quality (ICIQ 07), MIT, Massachusetts, USA, November 9-11, 2007. [24] A. Charnes, W. W. W. Cooper and L. Rhodes, "Measuring the efficiency of decision making units," European Journal of Operational Research, vol. 2, no. 6, pp. 429-444, November 1978. [25] C. Batini and M. Scannapieco, Data Quality: Concepts, Methodologies and Techniques (Data-Centric Systems and Applications), Secaucus,  NJ: Springer-Verlag Springer-Verlag New York, Inc, 200 2006. 6.

164