An adaptive optimal clustering number algorithm for FKCM

In order to overcome the drawbacks of fuzzy kernel clustering method (FKCM) give the clustering number in advance, sensitive to the initial cluster centers and easy to be trapped into local optimum, the adaptive algorithm for optimal clustering number of FKCM (SAICFKCM) is proposed. The proposed method uses density-based algorithm to initialize cluster centers and kernel Xie-Beni validity index to determine the optimal number of categories, to achieve unsupervised fuzzy partition of data set. The simulation experiment and the classification of naphtha attribute data verify the feasibility and effectiveness of the proposed method.