基于二次插值的GM(1,1)模型预测方法的改进

The Improvements of Forecasting Method in GM(1,1) Model Based on Quadratic Interpolation

从理论上分析了GM(1,1)模型中背景值的计算,指出文[11]利用Newton-Cotes公式构造模型背景值的方法是不可靠的,提出用二次插值构造模型中的背景值,同时用最小二乘法对预测公式中的初值进行改进,并用改进的方法进行了短期预测。其理论分析及仿真结果均表明本文所提出的方法有效可靠,为提高预测精度提供了新的途径。

In this paper, we analyse the background value calculation in GM（1,1） model , and point out that the structure method of background value using the Newton-Cotes formula in [11] is irreliable, and propose a new method to build the background value based on quadratic interpolation, at the same time, by using the least square method we imporve the initial value in the forecasting formula, and conduct a short-term prediction by the im- proving method. Both the analysis in theory and the simulation result show the effectiveness of the proposed approach, and give a new way to improve the prediction accuracy.