A fuzzy cluster algorithm based on mutative scale chaos optimization

Abstract

The traditional Fuzzy C-Means (FCM) algorithm has some disadvantages in optimization method, which makes the algorithm liable to fall into local optimum, thus failing to get the optimal clustering results. According to the defect of FCM algorithm, a new Fuzzy Clustering algorithm based on Chaos Optimization (FCCO) is proposed in this paper, which combines mutative scale chaos optimization strategy and gradient method together. Moreover, a fuzzy cluster validity index (PBMF) is introduced to make the FCCO algorithm capable of clustering automatically. Three other fuzzy cluster validity indices, namely XB, PC and PE, are utilized to compare the performances of FCCO, FCM and another algorithm, when applied to artificial and real data sets classification. Experiment results show FCCO algorithm is more likely to obtain the global optimum and achieve better performances on validity indices than other algorithms. © 2008 Springer-Verlag Berlin Heidelberg.