混沌时间序列预测的建模与仿真研究

摘要

研究混沌时间序列预测准确性问题,由于混沌时间序列具有混沌性和非线性,传统时间序列预测方法不能准确将混沌时间序列变化规律计算出来,导致预测精度低.为了提高混沌时间序列预测的精度,提出一种改进支持向量机的混沌时间序列预测方法(PSO-LSSVM).PSO-LSSVM采用相空间重构对混沌时间序列进行重构,去除其混沌性,用支持向量机对非线性进行预测,并采用粒子群算法对支持向量机参数进行优化,对经典混沌时间序列Mackey-Glass最优模型进行仿真测试.仿真结果表明,PSO-LSSVM加快了预测速度,提高了预测精度,在混沌时间序列预测中具有很好的应用价值.%Chaos time series prediction is currently a hot issue, the prediction precision of conventional time series prediction methods is very low, and a chaotic time series prediction method is put forward based on particle swarm optimization and least squares support vector machine(PSO-LSSVM).Firstly, PSO-LSSVM uses phase space reconstruction to reconstructthe sample, uses particle swarm algorithm to optimize the LSSSVM parameters, and gets the optimal chaotic time series prediction model.Finally the optimum model is tested and analysised by MackeyGlass chaotic time series.Simulation results show that the PSO-LSSVM accelerates prediction speed and improves the peidiction accuracy compared with the other prediction methods, and PSO-LSSVM has a good application prospect in chaotic time series prediction.