时序建模中周期信号和趋势项的处理问题研究

STUDY ON THE PROCESSING NETHOD OF PERIODIC AND TENDENTIOUS SIGNAL IN TIME SERIES MODEL -BUILDING

利用格林函数的概念，探讨了自回归滑动平均模型（ＡＲＭＡ）对周期信号和多项式趋势信号的描述问题。笔者通过理论推导和建模实作，指出ＡＲＭＡ模型或ＡＲ模型可以正确表述周期性信号和趋势性信号，因此在时间序列建模时，不必要也不应该将上述成分剔除。这一观点有别于对该问题的传统认识，并进一步探讨了造成这一错识的原因。

By using the Green Function formula of the autoregressive-move-averagemodel(A RMA model),the describing problem of the periodic and tendentioussignal iS discussed. Based on theory deduction and modelling practice,it isobtained that the above signals can be correctly expressed by an ARMA model orAR model.So in the model-building process of time series, it is unnecessary andimproper to eliminate these signals.This viewpoint differs from the traditionalone of this problem. The reason why the wrong view was developed is also discussed in this paper.