相空间重构和支持向量机参数联合优化研究

Jointly optimization for parameters SVM and phase space reconstruction

在混沌时间序列预测过程中,相空间重构和支持向量机参数是影响混沌时间序列预测性能的两个重要方面,传统上两者是分开单独进行的.利用相空间重构和支持向量机参数之间的互相依赖关系,提出了一种基于粒子群算法的相空间重构和支持向量机参数联合优化方法.参数联合优化核心思想是在相空间重构的同时选择最优支持向量机参数,通过粒子群算法对参数联合优化来实现.通过采用参数联合优化算法对混沌时间序列Mackey-Glass和太阳黑子年平均数时间序列进行了仿真实验,结果表明,相对于传统的分开单独优化方法,参数联合优化方法提高了混沌时间序列模型的预测精度,泛化能力更好.

Phase space reconstruction and support vector machine parameters optimization are two important aspects in chaotic time series prediction which are solved separately traditionally.A method was developed for jointly optimization of phase space reconstruction and support vector machine parameters,using the interdependent relationship between phase space reconstruction and support vector machine parameters to improve the model prediction performance.The main idea of the parameters jointly optimization method was that searching support vector machine parameters during phase space reconstruction by particle swarm optimization.Simulation experiments were carried out on the Mackey-Glass and the sunspot number chaotic time series,the results show that the parameters jointly optimization method improves the prediction accuracy and run more efficiently.