A Novel Clustering Based Classifier Using Support Vector Machines Criterion

In this paper, a novel clustering-based classifier using Support Vector Machines criterion (called CBCSVM) is presented for pattern classification. This algorithm involves three steps. At first, the robust clustering algorithm Kernelized Fuzzy c-means is utilized to yield the clustering centers. Then, a set of Gaussian functions associated with these obtained centers are adopted to map the samples to a new feature space to enhance the separability among different classes. Finally, the SVM criterion is applied in the transformed feature space to complete the classification. This algorithm has two advantages: (1) By mapping the samples into a new feature space, the separability among different classes is possibly enhanced according to the Cover´s theorem. (2) By inducing the robust clustering information into classification process, the prior information about the structure distribution is incorporated into the classification process and thus the classification performance is improved. The experiments on the benchmark datasets demonstrate that the proposed algorithm works better than some classical algorithm such as Radial Basis Function neural network and SVM.