Temperature Forecast Based on SVM Optimized by PSO Algorithm

This paper presents a new SVM algorithm framework optimized by PSO algorithm. The value of parameters in the SVM has great influence on the performance of regression model. In previous works the choice of these parameters mainly depends on the experience. In our work PSO algorithm was used to optimize these parameters to form a new SVM framework - PSVM. The proposed algorithm was used to forecast the daily minimum temperate. The experimental results show that the proposed PSVM has higher accuracy the some other SVM model such as GSVM and basic SVM.