摘要
给出求解组合预测权系数的回归分析方法 ,文章首先给出了基于最小二乘和最小一乘准则的线性回归组合预测模型 ,然后应用最小二乘原理得到权系数最小二乘估计值。由于最小一乘准则下 ,目标函数不可微 ,传统的优化规划方法无法求解 ,故文中提出用基于最小二乘的逐步变权方法进行求解。同时 ,还给出了百分误差绝对值最小为目标的组合预测模型及权系数求解方法。通过实例分析 ,表明组合预测模型的预测精度很高 。
Using regression analysis method,the methods for solving the weights of combination forecasting model(CFM) are proposed. At first, the linear regress CFM are presented based on the least absolute criteria and least square criteria. Then the weights can be evaluated using the least square princinple. Because the objective function of CFM based on least absolute criteria is non differential, the traditional programming methods can not solve it. So the least square method with the modified weights is proposed to solve this problem. At the same time, methods for solving CFM is given with the aim of minimizing sum of percentage error absolutes. From many cases, the results show that the forecasting precision of CFM is very high and the effect of regression is remarkable.
出处
《重庆大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2003年第1期62-65,共4页
Journal of Chongqing University
基金
重庆大学青年骨干教师资助基金
