Graph approximation on similarity based rough sets

Abstract

Correlation clustering is a widely used technique in data mining. The clusters contain objects, which are typically similar to one another and different from objects from other groups. In the authors previous works the possible usage of correlation in rough set theory were investigated. In rough set theory, two objects are treated as indiscernible if all of their attribute values are the same. A base set contains those objects that are indiscernible from one another. The partition, gained from the correlation clustering, can be understood as the system of base sets, as the clusters contain the typically similar objects (not just to a distinguished member) and it considers the real similarity among the objects. In this work the extension of this study is presented, using the method to approximate graphs representing similarity relations.