基于L1范数的加权几何平均组合预测模型的性质

提出了基于L1范数的加权几何平均的组合预测模型,针对该模型定义了优性组合预测、预测方法优超和组合预测冗余度等新概念;指出模型的任一个可行解对应的组合预测至少是非劣性组合预测, 探讨简单等权平均组合预测方法为最优性组合预测方法和优性组合预测存在的充分条件;论证预测冗余信息的可能存在性并给出冗余预测方法出现的一个判定定理.最后进行了实例分析,结果表明该方法是一种有效的组合预测方法.

Properties of weighted geometric means combination forecasting model based on L1 norm

Based on L 1 norm, a weighted geometric means combination forecasting model is proposed. Some new concepts such as superior combination forecasting, dominant forecasting method, combination forecasting redundant degree are defined for the model. It is pointed out that combination forecasti ng corresponding to the arbitrary feasible solution of this model is at least non inferior. The sufficient condition that equally weighted geometric means comb ining forecasting is optimal and existence condition of superior combination forecasting are discussed. Redundant forecasting information is probably existent in this model and its determining theorem is proved. Finally a numerical example is illustrated to show that this model is an effective combination forecasting method.