Parameters selection and application of support vector machines based on particle swarm optimization algorithm
基于粒子群优化算法的支持向量机参数选择及其应用

Parameters selection is an important problem in the research area of support vector machines (SVM), and its nature is an optimization problem. Motivated by the effectiveness of evolution algorithm on optimization problem, a new automatic searching methodology, based on particle swarm optimization (PSO) algorithm, is proposed in this paper. Each particle indicates a group of SVM parameters, and the population is a collection of particles in this method. Furthermore, the k-fold cross-validation error is used as the fitness function of PSO. After having been validated its effectiveness by two artificial data experiments, the proposed method is then applied to establish a soft-sensor model for average molecular weight in polyacrylonitrile productive process. Finally, real data simulation results are also given to show the efficiency.