组合预测方法预测误差平方和上界的几个新结果

Some New Results of the Upper Bound of the Sum of Squares of Combination Forecasting Error

本文给出一个更强且更为一般的结果,即组合预测方法预测误差平方和小于各个单项预测方法预测误差平方和的加权平均值,且必然小于各个单项预测方法预测误差平方和的最大值。并证明了简单加权平均法和方差倒数加权平均法预测误差平方和的上界均小于简单平均法预测误差平方和的上界。

In this paper, we study the upper bound of the sum of squares of combination forecasting error and obtain some new results. We prove that the sum of squares of combination forecasting error is less than the weighted average of the sum of squares of forecasting error of all the individual forecasting methods, and necessarily less than the maximum of the sum of squares of forecasting error of all the individual forecasting methods.