Using Information Gain in Data Fusion and Ranking

Abstract

Entropy and information gain have been traditionally used to measure association between inputs and outputs. In this paper, Information gain is used to measure and decide the level of dependency or relevance between attributes. A data fusion technique based on information gain measures in a virtual data integration environment is introduced. After the detection and clustering of duplicates, the fused records are ranked and provided to the user in the final answer set with a preference score associated with each answer