Abstract
This paper is concerned with a comparative study of the performance of fuzzy clustering algorithms Fuzzy C-Means (FCM), Gustafson-Kessel FCM (GK-FCM) and two variations of kernel-based FCM. One kernel-based FCM (KFCM) retains prototypes in the input space while the other (MKFCM) implicitly retains prototypes in the feature space. The two performance criteria used in the evaluation of the clustering algorithm deal with produced classification rate and reconstruction error. We experimentally demonstrate that the kernel-based FCM algorithms do not produce significant improvement over standard FCM for most data sets under investigation It is shown that the kernel-based FCM algorithms appear to be highly sensitive to the selection of the values of the kernel parameters.