逐步回归周期分析与多层递阶方法相结合的时间序列分析预报模型

A TIME SERIES FORECOSTING METHOD COMBINING THE STEPWISE REGRESSION PERIOD ANALYSIS WITH THE RECURSIVE ESTIMATE METHOD

摘要: 逐步回归周期分析1是针对方差分析的不足而提出的一种时间序列分析方法,作者2将其应用于长期预报,取得了一定的效果,但也存在着某些缺陷,其主要表现在实际应用时,随着预报时效的增长,预报效果明显趋于不稳定.这里主要有两个方面的原因:一是某些时间序列本身的周期性不强,所选的周期显著性水平低,此时则不宜单纯用周期分析外推做长期预报,对此,作者已在另文中作过研究.二是在逐步回归周期分析预报模型中,回归系数取为固定的参数,而天气系统则是时变系统.

Abstract: The paper is aimed at the defect of the stepwise regression period analysis,a improved time series forecasting model is presented.It combines the stepwise regression period analysis with the recursive estimate method,and isapply ed to the long-range forecasting of the precipitation on June at Jinhua City.The result shows that the improved method is satisfying.