基于支持向量机的混沌时间序列预测分析

Prediction analysis of chaotic time series based on support vector machines

混沌的特性决定了混沌系统很难长期预测,支持向量机有强大的学习能力,根据相空间重构理论用支持向量机建立预测模型对混沌时间序列进行短期预测。预测输出构建混沌吸引子来定性评价预测模型性能,同时与BP神经网络RBF神经网络构建的预测模型比较,计算预测模型的均方根误差定量地评价模型的性能。仿真结果表明,该方法具有较高的预测精度和泛化能力。

It is known that long-term prediction of chaotic time series is not possible for the nature of chaotic systems.Support vector machines（SVM） have powerful learning ability.The prediction model of support vector machines in combination with phase reconstruction of chaotic time series had been established,which was used in short term prediction of chaotic time series.A qualitative evaluation of the prediction models was made with the reconstruction of an attractor by the prediction models,and a quantitative evaluation of the prediction models was made with calculation of the root mean square error of prediction models output.This paper applied support vector machines to chaotic time series prediction,and compared the results of prediction with BP network and RBF network.The results of simulation experiments show that SVM has high prediction accuracy and generalization ability.