Cohesion based attribute value matching

In relational data, identifying the distinct attribute values that refer to the same real-world entities is an essential task for many data cleaning and mining applications (e.g., duplicate record detection and functional dependency mining). Existing approaches for attribute value matching include string-based approaches and rule-based approaches. However, the effectiveness of string similarity based approaches may severely rely on the choice of string similarity metric, proper threshold and even domain knowledge, while rule-based approaches suffer from the neighbour sparsity problem. To address such limitations, a cohesion-based metric is proposed to improve the matching performance between distinct attribute values. Additionally both string similarity and cohesion-based measurement are combined into a unified ranking by an evidential reasoning approach. Experimental evaluations show that our approach can improve the performance of matching attribute values greatly.