A New SVM Merged into Data Information

A new technique using a positive symmetric function to improve support vector machine (SVM) is presented. Firstly, the support vectors are obtained from traditional SVM. Secondly, a positive scalar function is constructed using the support vectors. Thirdly, a new kernel function is obtained from the Gaussian kernel multiplied by the positive symmetric function merged into data information. The new SVM with the new kernel is to enlarge margin around the separating hyper-plane as a result of the differential approximation theory. Therefore, the separability is increased and the support vectors are decreased. Simulation results for both artificial and real data show remarkable improvement of generalization error and computational cost.