Approximating Source Accuracy Using Dublicate Records in Data Integration

Abstract

Currently, there are two main basic strategies to resolve conflicts in data integration: Instance-based strategy and metadata-based strategy. However, the two strategies have their limitations and their problems. In a metadata-basedstrategy, the data integration administrator uses some features of source to determine how they may participate to provide a single consistent result for any conflicted object. One of the most important featuresin metadata-based strategyis source accuracy, which is concerned with the correctness and precision with which real world data of interest to an application domain is represented correctly in the data store and used for conflict resolution. However, it is difficult and time consuming to calculate accuracy for all data sources and compare data about objects with what is in real world, in this article we propose a new way to approximate sources accuracy using known accuracy of predefined source as a reference. In this paper wepresenthow to determine most appropriate source as a reference using some criteria do not depend on comparing with real objects. After that we demonstrate how to use this reference source and its known accuracy to approximate other sources accuracies using common objects between them and reference source.