Credibility Coefficients for Objects of Rough Sets

Abstract

In this paper focus is set on data reliability. We propose a few methods, which calculate credibility coefficients for objects stored in decision tables. Credibility coefficient of object is a measure of its similarity with respect to the rest of the objects in the considered decision table. It can be very useful in detecting either corrupted data or abnormal and distinctive situations. It is assumed that the proper data appear in majority and can be separated from improper data by exploring mutual resemblance. The proposed methods take advantage of well known and widely used data mining technique - rough sets.