基于改进RBF神经网络的混沌时间序列预测  

Chaotic Time Series Forecasting Model Based on the Improved RBFNN

作　　者：郭兰平[1] 俞建宁[1] 张旭东[1] 漆玉娟[1] 张建刚[1] 

机构地区：[1]兰州交通大学数理与软件工程学院,甘肃兰州730070

出　　处：《云南民族大学学报（自然科学版）》2011年第1期63-70,共8页Journal of Yunnan Minzu University:Natural Sciences Edition

基　　金：国家自然科学基金(408034);甘肃省自然科学基金(3ZS051-A25-030;3ZS-042-B25-049);兰州交通大学大学生科技创新基金(DXS2010-021)

摘　　要：基于RBF神经网络与相空间重构理论,对网络预测模型进行改进,并以Lorenz动力系统产生的混沌时间序列作为研究对象,建立预测模型并对其进行数值仿真.实验结果表明,基于改进RBF神经网络与相空间重构理论的混沌时间序列预测方法比BP、RBF神经网络模型的预测精度高、误差小、性能优越,改进方法可行、有效.Based on RBF neural network and theory of phase space reconstruction, this research improved the net- work prediction model, and took the chaotic time series that was generated by Lorenz dynamical system as the research object. Prediction models were built and their numerical simulation was carried out. Experimental results show that with the improved RBF neural network and the theory of phase - space reconstruction of chaotic time series forecasting model with higher predictive precision, this one has fewer errors and superior performance than the BP or RBF neural network model, and so it is feasible and effective.

关 键 词：RBF神经网络 相空间重构 嵌入维数 延迟时间 混沌时间序列 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]