季节性时间序列预测方法选择

Time series forecasting methods selection for seasonal data

现实生活中大部分的经济数据不仅会随着时间的推移显示出一定的长期趋势,往往还会因为季节性因素而呈现出周期变化,因此,对于这种既具有倾向性变动趋势又有季节性变动的时间序列的预测就成为了统计预测的重要内容之一。因为预测方法选择的多样性,主要讨论温特线性与季节性指数平滑法,自适应过滤法和ARIMA模型拟合法这3种重要且比较典型的预测方法,通过比较3种方法的优劣,有助于在实际预测中预测方法的正确选择。

In reality,most economic data will show certain long-term trends as time goes by,and exhibit periodical changes affected by seasonal factors.As an important focus of statistical prediction,it is very necessary to study and predict such time series with tendency and seasonality.Although there are many statistical prediction methods available,this paper mainly discusses three important and typical ones,Winters linear and seasonal exponential smoothing,adaptive filtering model,and ARIMA model.The purpose of this paper is to help choose the right method by comparing advantages and disadvantages of these methods in actual predictions.