Comparative study of a genetic fuzzy c-means algorithm and a validity guided fuzzy c-means algorithm for locating clusters in noisy data

Abstract

The partitioning of data into clusters is an important problem with many applications. Typically, one locates partitions using an iterative fuzzy c-means algorithm of one form or another. Unfortunately, the results of these techniques depend on the cluster center initialization because their search is based on hill climbing methods. Recently, there has been much investigation into the use of genetic algorithms to partition data into fuzzy clusters. Genetic algorithms are less sensitive to initial conditions due to the stochastic nature of their search. In this paper we compare the two techniques when locating fuzzy clusters embedded in noisy data and discuss the advantages and disadvantages of both methods.