Fuzzy Clustering for Categorical Data by Some Kinds of Fuzzifying Different from Fuzzy k-Partitions

Abstract

In fuzzy clustering algorithms for vectorial data, various types of fuzzification technique are utilized, whereas only Bezdek-type fuzzification has been applied for nominal data resulting in Fuzzy k- partitions (BFKPn). Furthermore, BFKPn may fail produce clusters with different sizes. In this report, three fuzzy clustering algorithms are proposed for nominal data, where all methods have variables control- ling cluster sizes. Furthermore, Bezdek-type fuzzification, entropy-regularization, and q-deformation are applied to each method for fuzzification.