传统的时间序列模型虽然可以通过差分形式来处理非平稳随机数据,但是当数据具有确定性效应的时候,对数据进行差分容易造成残差信息的浪费.通过对拟合后的残差建立AR模型,可以充分利用数据的趋势信息和残差信息,从而得到较好的预测效果.%Although the ARIMA model can process the non-stationary random data by calculus of differences, processing data with certain tendency through calculus of differences can easily lead to waste of residual information. By using the residual data after fitting the data curve, we can make full use of the trend information and the residual information. So we can get better prediction results.
展开▼