基于局域波法的时间序列预测

提出了一种新的用于时间序列预测方法。由于实际信号常常具有非平稳特征，直接应用AR模型进行时间序列分析，得不到理想的效果。而局域波分解法是一种新的分析非平稳、非线性的工具。用局域波分解法对待预测的时间序列进行经验模式分解，产生多个基本模式分量，对各个分量分别进行AR模型预测，然后重构各个预测值。仿真结果表明预测结果比直接进行AR模型预测更加精确。

Time series prediction with local wave method

The article proposes a new method for time series. The real signals have often non-stationary characteristic, so if we analyse these time series using AR model directly, we cant obtain design result. The local wave decomposition method is a new method to analyse instability and nonlinearity. First we decompose the time series analyzed using local wave decomposition, generating a few intrinsic mode functions, then predict all the intrinsic mode functions using AR mode respectively, lastly reconstruct the predicted value. Simulations suggest that this method can predict more accurately than that using the AR model directly.