Possibilistic Clustering in Feature Space
WILF '07 Proceedings of the 7th international workshop on Fuzzy Logic and Applications: Applications of Fuzzy Sets Theory
International Journal of Computational Intelligence Studies
Hi-index | 0.00 |
A fuzzy clustering method is presented based on kernel methods. The proposed model is called kernel possibilistic fuzzy c-means model (KPFCM). It is claimed that KPFCM is an extension of possibilistic fuzzy c-means model (PFCM) which is superior to fuzzy c-means (FCM) model. Different from PFCM and FCM which are based on Euclidean distance, the proposed model is based on non-Euclidean distance by using kernel methods. Furthermore, with kernel methods the input data can be mapped implicitly into a high-dimensional feature space where the nonlinear pattern now appears linear. KPFCM can deal with noises or outliers better than PFCM. The proposed model is interesting and provides good solution. The experimental results show better performance of KPFCM.