Study Of Weighting Fuzzy Clustering Algorithm Based On Generalized Entropy

Entropy fuzzy clustering is a combination method of fuzzy clustering and entropy. It hasnot only the advantages of entropy in expressing related information among the data samplesbut also the characteristics of quality about soft clustering in fuzzy clustering algorithms, so itoccupies an important place in the field of data clustering. Through the analyses of entropyfuzzy clustering algorithm, further research on fuzzy clustering method by combininggeneralized entropy, the weights of samples and Kernel function is done in this paper. Thespecific contents are as follows:Through weighting the data samples and combining it with generalized entropy fuzzyclustering method, the objective function of weighted generalized entropy fuzzy clusteringmethod is obtained; Further more, this paper gives the weighted generalized entropy fuzzyclustering algorithm and its iterative calculation formula of subjection degree and clustercenter which are received by using the lagrangian method. In addition, research on how todetermine the weights of data samples in the weighted generalized entropy fuzzy clusteringmethod is also done.On the basis of weighted generalized entropy fuzzy clustering method, throughintroducing the kernel function we obtain the objective function of kernel weightedgeneralized entropy fuzzy clustering method and give the iterative calculation formula ofsubjection degree and cluster center in theory. Then we further present the kernel weightedgeneralized entropy fuzzy clustering algorithm. In addition, the problems of how to combineor structure the kernel function which is used in the kernel weighted generalized entropyfuzzy clustering algorithm are studied. Then we can use the better kernel functions which aremore suitable for the characteristics of data sets in the kernel weighted generalized entropyfuzzy clustering method to improve the clustering effect more effectively.By selecting the representative data sets in cluster analysis field, this paper doesexperimental study on both the weighted generalized entropy fuzzy clustering algorithm andthe corresponded kernel algorithm, and then compare it to the traditional generalized entropyfuzzy clustering method. The experiment results show that the algorithms proposed in thispaper are effective.