一种基于最小二乘法的广义加权组合预测模型

A Combined Forecasting Model Based On Optimal Weighted Method And East-squares Method

研究数据预测问题,传统最小二乘法所确定的单项预测模型,一方面不能很好地利用不同模型的信息,另一方面当拟合效果不佳时预测误差较大。针对最小二乘法的缺点,构建了一种新的预测模型———广义加权最小二乘组合预测模型。通过最小二乘法确立多个单项预测模型,经过广义加权组合法综合不同模型的信息,利用非线性规划法求解最优权系数,最终得到组合预测模型。经仿真计算,证明了该方法的可行性及优越性。

Studying about the data forecasting problem. The least-squares method can＇t make a full use of the informa- tion of different models, and will reduce the forecast precision since the fitting results is not ideal. According to the weakness, a new forecasting model based on generalized weighted proportional means and least-squares method is estab- lished. Firstly, the model used the least-squares method to determine the single forecast models. Secondly, the model adopted the generalized weighted proportional means to syncretize different informations from the single. Thirdly, the model estimate optional weight coefficients through nonlinear programming and the combined forecasting model can be get.The results of simulation demonstrated that the combined forecasting model was better than a single method.