组合预测方法研究的若干新结果

Some New Results of Combination Forecasting Study

本文对组合预测方法的一些基本问题进行了较深入的研究:证明了随着预测方法的增加,最优组合预测方法的预测误差平方和并不一定减少;给出了最优组合预测方法的预测误差平方和恰好等于参加组合预测的各个单项预测方法预测误差平方和的最小值的充分条件;对递归等权组合预测方法的收敛性等问题给出了理论证明,并给出了一种新的替换准则。本文还就最优组合预测方法预测误差平方和的取值范围等问题进行了研究,给出了一些新的结果。

In this Paper,we study several theoretic problems of combination forecasting method, We prove that the sum of square error of optimal combination forecasting method does not ncesserily decreasesas the number of forecasting methods increases, and that the sum of square error of optimal combination forecasting method is not necessarily less than that of any of the forecasting methods to be combined. We also prove that the Recursive Equal Weighting method is convergent and more efficient than simple average forecasting method.