In order to improve the generalization performance and prediction accuracy of LSSVM based time series prediction,a PSO based LSSVM was studied.Firstly,a certain number of LSSVMs were trained by using training samples and then cross-validation error was applied to evaluate the generalization performance of the LSSVMs.Finally,PSO was applied to search for the optimal LSSVM with the smallest cross-validation error.Experiments on time series prediction indicate that LSSVM optimized by PSO has better prediction performance than that not optimized and conventional prediction methods.%为提高基于最小二乘支持向量机(LSSVM)的时间序列预测方法的泛化能力与预测精度,研究了一种基于粒子群优化(PSO)的LSSVM。该方法以交叉验证误差为评价准则,利用PSO对多个具有不同超参数的LSSVM进行基于迭代进化的优化选择,并以交叉验证误差最小的LSSVM作为最终优化后的LSSVM。时间序列预测实例表明,经PSO优化后的LSSVM的预测精度高于未经优化的LSSVM与传统时间序列预测方法的预测精度。
展开▼