混沌时间序列基于邻域点的非线性多步自适应预测  

Nonlinear adaptive multi-step-prediction of chaotic time series based on points in the neighborhood

作　　者：甘建超[1,2] 肖先赐[1] 

机构地区：[1]电子科技大学电子工程学院 [2]电子对抗国防科技重点实验室,成都610036

出　　处：《物理学报》2003年第12期2995-3001,共7页Acta Physica Sinica

基　　金：国防科技预研基金 (批准号 :5 14 3 5 0 5 0 10 1DZ0 2 0 3 )资助的课题~~

摘　　要：根据流形理论 ,利用混沌时间序列中某点邻域内最近几点的P次迭代像 ,提出了一种多步自适应预测算法 .仿真说明 ,这种算法使得预测速度成倍提高 ,而预测稳定后得到的误差均方根序列呈指数增长趋势 ,这个指数就是该混沌时间序列的Lyapunov指数 .In this paper a class of nonlinear adaptive multi-step-prediction algorithm based on the manifold theory was proposed. We have performed the multi-step-prediction by exploiting images of P-step iterations of several nearest neighbors with this method. The simulation indicated that this method was available and could improve the prediction speed, and that the series of the standard deviation of error after prediction has an exponential growth ratio that is the largest Lyapunov exponent.

关 键 词：混沌时间序列 邻域 非线性自适应预测 LYAPUNOV指数 S-M算法 预测模型 嵌入维数 

分 类 号：O415.5[理学—理论物理]