Fuzzy rough clustering for categorical data

Abstract

Unlabeled categorical data is common in many applications. Because there is no geometric structure for categorical data, how to discover knowledge and patterns from unlabeled categorical data is an important problem. In this paper, a fuzzy rough clustering algorithm for categorical data is proposed. The proposed algorithm uses the partition of each attribute to calculate the granularity of each attribute and introduces information granularity to measure the significance of each attribute. It is different from traditional clustering algorithms for categorical data that the proposed algorithm can transform categorical data set into numeric data set and introduces a nonlinear dimension reduction algorithm to decrease the dimensions of data set. The proposed algorithm and the comparison algorithms are executed on real data sets. The experimental results show that the proposed algorithm outperforms the comparison algorithms on the most data sets and the results prove that the proposed algorithm is an effective clustering algorithm for categorical data sets.