Abstract

Fuzzy c-means (FCM) clustering is a well-known fuzzy clustering algorithm and possibilistic c-means (PCM) clustering is a good model-seeking algorithm. However, FCM is sensitive to noises or outliers and PCM is very sensitive to initializations and sometimes generates coincident clusters. To overcome their shortcoming, two novel fuzzy clustering algorithms, called united c-means (UCM) clustering algorithms, are proposed in this paper. UCM integrates FCM and PCM. The membership values from UCM are the product of the membership values from FCM and the typicality values from PCM. Experiments show the better performances of UCM.

Paper Details