A modified possibilistic fuzzy c-means clustering algorithm

Possibilistic clustering algorithm can give the fuzzy and possibilistic partition of the data set. This paper analyzes the mean shift clustering algorithm (MSC) and the possibilistic fuzzy c-means clustering algorithm (PFCM) in detail, base on which a modified possibilistic fuzzy c-means clustering algorithm (MPFCM) is proposed. The analysis shows that PFCM has the initialization sensitivity problems, while MSC can determine the cluster number in different scales and it is independent to the initializations. MPFCM not only inherits the merit of both the PFCM and MSC, but also avoids the problems from them. The experimental results show the relatively better performance of the proposed algorithm on computation and initialization.