Model selection of support vector regression using particle swarm optimization algorithm

A new automatic search methodology for model selection of support vector regression (SVR), based on the particle swarm optimization (PSO) algorithm, was proposed to search for the adequate hyper-parameters of SVR. In this method, each particle indicates a group of hyper-parameters, and the population is a collection of particles. Two artificial data experiments results show that our method performs Superiorly oil function approximation. Furthermore, the proposed method was applied to establish a soft-sensor model for average molecular weight in polyacrylonitrile productive process. The results of real data simulation also show that this method is effective.