基于GM(1,1)模型和线性回归的组合预测新方法

A New Forecasting Model Based on the Combination of GM(1,1) Model and Linear Regression

为解决 GM(1 ,1 )预测中存在的历史数据的跳变问题 ,依据灰色灾变预测原理 ,利用线性回归适用短期预测的特点 ,提出了一种新的预测方法 :用 GM(1 ,1 )模型预测将来可能的数据跳变日期点 ,对其他非跳变点使用分段线性回归函数进行预测 .通过对浙江省农村用电量的预测 ,结果表明该方法很好地克服了 GM(1 ,1 )模型和线性回归模型的缺陷 ,在实际中取得了较好的效果 .

Aiming at solving the problem of the aberrant points' presence of the history data in GM(1,1) forecasting model,depending on the grey disaster forecasting theory and using the character of linear regression adapting to the short term forecast,a new combined method was come up with. After using GM(1,1) to forecast the possible aberrant date points in the future,the prediction values of the points were made. And for other normal points,subsection linear regression functions were taken. By having applied the new method to the prediction of the country power consumption in Zhejiang province,it showed that the new method had achieved better forecasting results compared with other forecasting models,making up for some deficiencies in GM(1,1) model and linear regression in certain degree.