A probabilistic cluster validity index for agglomerative bayesian fuzzy clustering

Abstract

A novel fuzzy clustering technique, called Agglomerative Iterative Bayesian Fuzzy Clustering(IBFC) with a novel cluster validity index is presented. The algorithm has a fuzzy competitive learning structure properly incorporated with Bayesian decision rule. Based on this Bayesian assumption, we propose a probabilistic cluster validity index, by which an optimal number of clusters is determined. We reports that the proposed algorithm shows better performance when tested with synthetic/benchmark data and compared with several well-known methods. © 2008 IEEE.