A KFCM-Based Fuzzy Classifier

A proposed KFCM-based fuzzy classifier was introduced. As for the process of constructing such classifier, firstly, the original sample space is mapped into a high dimensional feature space by selecting appropriate kernel function. Then in the feature space, training samples of each class are divided into some clusters by proposed KFCM algorithm. The optimal cluster number of every class is selected by our proposed method. For each created cluster, a fuzzy rule is defined with ellipsoidal regions. Finally, fuzzy classification rules are adjusted by GAs. The accuracy of the constructed classifier by our proposed methods are comparable to the maximum accuracy of the multilayer neural network classifier and the relational reference, and the training time is much shorter.