A fuzzy support vector machine algorithm for classification based on a novel PIM fuzzy clustering method

The support vector machine (SVM) has provided excellent performance and has been widely used in real-world classification problems. Fuzzy methods used on the SVM solve the problem that the SVM is sensitive to the outliers or noises in the training set. In this paper, a novel partition index maximization (PIM) clustering method is studied to get a more reasonable and robust fuzzy membership for fuzzy SVM (FSVM). First, we improve the PIM clustering algorithm to cluster each of the two classes from the training set to get proper data centers. Then an algorithm is given to modify the boundary of PIM and form a new training set with fuzzy membership degrees. Finally, we use FSVM to induce the final decision function to show classification results. All the results indicate that the performance of PIM-FSVM is excellent.