Investigation of influences of temperature and q value in fuzzy c-means method maximized with Tsallis entropy

Abstract

Fuzzy c-means maximized with Tsallis entropy that is a q extension of Shannon entropy has been proposed. However, how to determine the q value is still not known. For this reason, influences of a temperature and the q value on the fuzzy c-means method maximized with Tsallis entropy combined with direct annealing are investigated. Based on the results, a new clustering method in which the temperature and the q value are changed simultaneously is proposed. Experimental results show that the proposed method is more effective than the conventional method.