The support vector machine (SVM) is a powerful tool for classification problems. Unfortunately, the training phase of the SVM is highly sensitive to the noises in the training set. Noises are inevitable in real-world applications. To overcome this problem, SVM was extended to Fuzzy SVM (FSVM) by assigning an appropriate fuzzy membership to each data point. However, suitable choosing of fuzzy memberships and an accurate model selection remind fundamental issues. In this paper, we propose a new method based on the optimization methods to simultaneously generate appropriate fuzzy membership and solve the model selection problem for SVM family in linear/nonlinear, separable/non separable classification problems. Both SVM and Least Square SVM (LSSVM) are included in the study. The fuzzy memberships are built based on dynamic class centers. The Firefly Algorithm (FA), a recently developed nature-inspired optimization algorithm, provides variation in the position of class centers by changing their attributes values. Hence, adjusting the place of class center can properly generate accurate fuzzy memberships to cope with both attribute and class noises. Also, through the process of generating fuzzy memberships, FA can choose the best parameters for the SVM family. A set of experiments is conducted on nine benchmarking data sets of UCI data base. The experiments results show the effectiveness of the proposed method in comparison with the seven well-known methods from SVM literature.

Keywords