A combination forecasting method based on model evaluation and selection from forecasting-model-base
Yan Hong-sen · Kongzhi yu juece · 2004
In order to avoid the pitfalls (such as incomplete information and NP hard problems) in the traditional procedure of model selecting and combining from a model base, a combination method based on model evaluation and selection from a forecasting-model-base is proposed, on the principle of personnel scrutinizing method. All models that fit for the current forecasting problem are chosen and a participating model set is composed. The degrees of satisfaction are assigned according to their forecasting precisions on different indices and endowed with their corresponding weights. An experiment shows that the method has a good forecasting efficiency.