基于粒子滤波的混沌时间序列局域多步预测

Particle filtering based local multi-step prediction for chaotic time series

对混沌时间序列进行预测研究具有重要的价值和实用性,例如,进行股票预测,降雨量预测,温度预测。混沌时间序列预测的难点在于其不确定性和多步预测的困难性。一般利用最小二乘法求解模型参数,从而对混沌时间序列进行局域预测,但是预测精度不是很高。为了提高局域线性预测的精度,提出基于粒子滤波(PF)的混沌时间序列局域多步预测法,利用粒子滤波进行参数优化得到更准确的优化模型进行多步预测。仿真实验结果表明,该方法的单步和多步预测效果明显得到了提升。

It has important value and practicability （such as stock forecasting, rainfall forecasting and temperature forecasting） to predict the chaotic time series. It is difficult to predict the chaotic time series due to its uncertainty and realization of multi-step prediction. The least square method is used to solve the model parameters and perform local prediction for the chaotic time series, but has low prediction accuracy. In order to improve the accuracy of local linear prediction, a local multi-step prediction method based on particle filtering （PF） is proposed for chaotic time series. The particle filtering is adopted to optimize the parameters to obtain more accurate optimization model for multi-step prediction. Simulation results show that the muhi-step and single-step prediction effects of this method are improved significantly.