An Improved PSO for Parameter Determination and Feature Selection of SVR and its Application in STLF
Dong-Xiao Niu and Ying-Chun Guo
A novel support vector regression (SVR) optimized by an improved particle swarm optimization (PSO) combined with simulated annealing algorithm (SA) was proposed. The optimization mechanism also combined the discrete PSO with the continuous-valued PSO to simultaneously optimize the input feature subset selection and the SVR kernel parameter setting. By incorporating with SA, the global searching capacity of the proposed model was enhanced. The improved SAPSO was used to optimize the parameters of SVR and select the input features simultaneously. Based on the operational data provided by a regional power grid in north China, the method was used in short-term load forecasting (STLF). The experimental results showed the proposed approach can correctly select the discriminating input features and compared to the PSO-SVR and the traditional SVR, the average time of the proposed method in the experimental process reduced and the forecasting accuracy increased respectively. So, the improved method is better than the other two models.
Keywords: Support vector regression (SVR), particle swarm optimization (PSO),simulated annealing (SA), parameter determination, feature selection, short-term load forecasting (STLF).