A novel composite-kernel based PCM clustering algorithm

Abstract

Possibilitic c-means (PCM) clustering is a fuzzy clustering algorithm, which overcomes the shortcomings of fuzzy c-means clustering (FCM) algorithm that is sensitive to noise. This paper proposes a composite-kernel possibilitic c-means clustering (CKPCM) algorithm based on the mercer kernel theorem for linearly inseparable data. The algorithm presented in this paper can obtain better clustering effect and higher clustering accuracy than the traditional PCM clustering algorithm, and can achieve better linear separability. Experimental results show the effectiveness of the proposed algorithm.