Improved Kernel-based Fuzzy Clustering Algorithm Based on Maximum Tsallis Entropy

This paper proposes an improved kernel-based fuzzy clustering algorithm based on maximum Tsallis entropy (MTE-IKFC). The algorithm is on the basis of the kernel-based fuzzy clustering algorithm to reduce the influence of data that doesn’t belong to the current cluster on the iterative solution of the corresponding clustering center, and suppress the convergence of different clustering centers when the regularization coefficient is small. In order to effectively reduce the influence of differences in sample distribution on the solution of clustering center, the weight based on relative density is introduced to make the clustering results more reasonable. Finally, Shannon entropy is extended to Tsallis entropy with a generalization parameter q, which further improves the performance of the algorithm. In the experimental part, compared with the other four clustering algorithms, the proposed algorithm shows superior clustering performance and the best robustness.