一种两阶段数量组合预测方法及实证

A Novel Two-Stage Combination Forecasting Algorithm and an Empirical Study

数量是决策中一个重要指标，对数量的掌握不但要了解现状而且要对其未来的发展趋势进行研究和预测。当前预测的方法很多，不同的预测对象适用不同的预测方法，方法的选择和应用是预测的一个难点和重点。本文首先梳理了现有数量预测方法的特点，然后设计了一个两阶段的组合预测方法，主要思路是采用常用预测方法，首先进行分阶段预测，考察各方法的拟合优度来筛选单项预测方法，确定单项预测方法后，选用组合预测提高预测精度。利用国家自然基金申请数据进行了实证研究，结果表明，该方法平均相对误差为0．04％，拟合精度较高，具有较好的应用价值。

Improving forecasting capacity and accuracy has always been the unswerving focus and pursuit of numerous scholars. A combined forecast model is more accurate than a single forecasting model. There is much literature about combination forecasting, which mainly focus on the research of new combination methods, especially on intelligent algorithm ; however,the precision research on the combination forecasting are mostly limited within samples,lacking practical value. In addition, under different forecasting environments, how to select single forecasting pattern and combination method in combination forecasting has rarely been involved. This paper set up the Two-Stage Combination Forecasting model and forecast the amount of applied for NSFC from 2001 to 2012. The research results show that the forecast values are accurate：the average relative error is only 0.04% ,so it can be usedin the application as one of the efficient methods.