Simultaneous section of parameters and features for SVM based on the differential evolution algorithm

To improve the efficiency of SVM's parameters selection and features selection, a simultaneous selection method based on the differential evolution algorithm (DE-SVM) was proposed. In the coding mode, the vectors of DE are divided into parameter dimensions and feature dimensions The parameter dimensions are used directly to select the parameters, and by 'integer-binary conversion' the feature dimensions are used to select the features Several numerical experiments on UCI benchmark datasets show the effectiveness of the proposed method. Compared with the parameters and features simultaneous selection method based on particle swarm optimization (PSO-SVM), the DE-SVM has higher efficiency and stronger ability of features selection.