基于相关系数的优性组合预测模型研究

Research on properties of superior combined forecasting models based on correlation coefficients

传统的组合预测模型是从改善某种拟合误差角度建立的.基于相关系数的组合预测模型有别于传统的组合预测模型,它是研究组合预测方法的一个新思路.针对该模型,提出了新的优性组合预测、预测方法优超,冗余度等概念;探讨了非劣性组合预测、优性组合预测以及冗余预测方法的存在性,并证明冗余预测方法的一个判定定理.这在理论上表明基于相关系数的组合预测方法的有效性,最后文中进行实例分析也验证了这个结论.

The traditional combined forecasting models are constructed by improving certain fitting errors. The combined forecasting model based on correlation coefficients is different from traditional one, which is a new idea to study combined forecasting methods. Some new concepts are put forward in this paper, such as superior combined forecasting, dominant forecasting method, redundant degree etc. Then some basic problems are discussed that are existence of noninferior combination forecasting, superior combination forecasting and redundant forecasting method. Finally a judgement theorem of redundant forecasting method is proved, which shows theoretically that this model is effective. This conclusion is checked by an illustrated numerical example.