基于一类损失函数的组合预测模型构建

Study on construction of combination forecast model based on loss function

组合预测的关键问题就是确定各单项预测的权系数,通过构造一类损失函数,以损失函数最小为目标来确立组合预测模型中各单项预测的权系数.比较了损失函数中的参数取不同值时对应的组合预测的预测效果,得出了在变权组合预测中,并非只有一组权系数能使预测效果达到最优.

The key problem of combination prediction is to determine the power coefficient of each single item prediction. By constructing a class of loss functions, the minimum loss func- tion as the target is the way to calculate the weight coefficient of each single prediction in the combination forecast model in this paper. And the forecast effects of the corresponding com- bination forecast when the parameters in loss function take different values were compared. In the variable weight combination forecast, it was not only a set of weights can make predic- tion effect to achieve the optimal.