基于多元局部多项式方法的混沌时间序列预测

Prediction of the chaotic time series using multivariate local polynomial regression

根据Takens定理,把混沌时间序列构造为一组序列对,然后用多元局部多项式方法来预测其序列.这种核估计方法可以结合局域法与全局法的优点,使得预测的精度更高.仿真结果表明,该方法非常有效.

According to the Takens theorem, we reconstruct the chaotic time series to multivariate data and use the multivariate local polynomial estimator to predict the chaotic time series. This kernel estimator has the advantages of both the local model and the global model. The simulation shows that the estimator can predict the chaotic time series accurately.