基于预报效果的ARIMA模型筛选

目的通过分析不同提前期的预报结果与提前量之间的关系,筛选出与预报目的相适应的ARIMA模型,从而使筛选得到的模型有可靠的预报效果.方法按照时间顺序逐段选择观察长度为50的序列片段并拟合ARIMA模型,初步筛选出6个在某些时段具有良好拟合效果的模型,借助这些模型进行提前期分别为1、2、5、10的预报,用对应分析方法考察预报效果与提前期间的关系.结果选用不同的模型,在不同提前期预报的精密度和精确度会有所不同.结论在以预报为目的的时间序列分析应用中,为了求得最终的良好预报效果,可以仿照本文提供的方法考核若干初选模型的预报特性,进而确定模型,而不是拘泥于拟合效果.这对ARIMA模型之外的其他形式预测模型的应用也有普遍指导意义.

Choose a Proper ARIMA Model Based on the Effectiveness of Prediction

Objective To choose a proper ARIMA model based on the relationship between the effectiveness of prediction and lead times.This makes the final models for prediction are of good quality.Methods Six ARIMA models are got by analysis of the sub-sequences that includes 50 observations respectively. Correspondence analysis is used to evaluate the relationship between the lead times, accuracy and precision of the predictions.Results Different model may be taken as the proper one as to different predicting consideration.Conclusion The method provided here is supposed to help the researcher to utilize a effective model to predict the future situation of a time series, based on the relationship between lead times and models, not stuck on the goodness of fit of a particular model.The research is also can be easily generalized to the prediction with models besides ARIMA.