A modified PCM clustering algorithm
Kai Li, Houkuan Huang, Kunlun Li · 2004
A fast PCM clustering algorithm is proposed in this paper. First, the fuzzy and possibilistic c-means (FCM and PCM ) clustering algorithms are analyzed and some drawbacks and limitations are pointed out. Second, based on the reformulation theorem, by means of modifying PCM model, an effective and efficient clustering algorithm is proposed here, which is referred to as a modified PCM clustering (MPCM). As eliminating the computation of membership parameters in each iteration, this algorithm saves an amount of running time. Finally, experiments are implemented by using MPCM clustering algorithm, and the chosen methods of parameters are also discussed. Experiments show that MPCM has not only abilities of resisting noise and avoiding trivial solution, but has fast clustering ability.