A projection and contraction method for circular cone programming support vector machines

The second-order cone programming support vector machine (SOCP-SVM) formulations have received much attention as the robust and efficient framework for classification. In this paper, we formulate the SOCP-SVM as the convex quadratic circular cone programming support vector machine (CCP-SVM). A projection and contraction method is used to solve the CCP-SVM. Experiments on the benchmark datasets from the UCI Repository and synthetic dataset show that the projection and contraction method for the CCP-SVM needs less computation time than the primal-dual interior point method (implemented by SeDuMi) for the SOCP-SVM. In addition, the proposed method has the almost similar accuracy, F-measure values and G-mean values as the primal-dual interior point method for the linear classifiers. The proposed method for kernel-based nonlinear classifiers can obtain higher performances of accuracy, F-measure and G-mean than the primal-dual interior point method for SOCP-SVM in some datasets.