组合预测方法中权系数的应用研究

Application Research on Weight Coefficients of the Combination Forecasting Method

针对组合预测方法中权系数的确定问题,提出两种组合预测方法。一种是引入误差评价公式来确定权系数的固定权系数组合预测方法;一种是基于神经网络的误差倒数变权组合预测方法。通过实例分析说明了两种方法的可行性,且结果表明方法比单模型方法预测和传统组合预测方法具有更高的拟合精度。

Aiming at the issue of determination of weight coefficient of combination forecasting method, two kinds of combination forecasting methods are presented. One is the stationary weight combination forecasting method whose weight coefficient is determined by introducing appropriate error criterion; the other is the error reciprocal variable weight combination forecasting method based on artificial neural network. The example shows that two kinds of methods proposed are feasible and better than single model prediction method and traditional combination forecasting method.