A Novel Hybrid Particle Swarm Optimization for Feature Selection and Kernel Optimization in Support Vector Regression

Abstract

This study proposed a novel HPSO-SVR model that hybridized the particle swarm optimization (PSO) and support vector regression (SVR) to improve the regression accuracy based on the type of kernel function and kernel parameter value optimization with a small and appropriate feature subset, which is then applied to forecast the monthly rainfall. This optimization mechanism combined the discrete PSO with the continuous-valued PSO to simultaneously optimize the input feature subset selection, the type of kernel function and the kernel parameter setting of SVR. The proposed model was tested at monthly rainfall forecasting in Guangxi, China. The results showed that the new HPSO-SVR model outperforms the previous models. Specifically, the new HPSO-SVR model can correctly select the discriminating input features, also successfully identify the optimal type of kernel function and all the optimal values of the parameters of SVR with the lowest prediction error values in rainfall forecasting.