摘要
针对灾害应急物资需求量时间序列的小样本和振荡性特征,提出了基于Fourier-GM(1,1)模型的应急物资需求量预测方法。该方法首先对给定的小样本振荡序列建立具有自适应背景值的GM(1,1)模型,然后应用Fourier级数描述模型残差中所包含的周期性振荡信息,进而构建Fourier-GM(1,1)模型。在此基础上,利用遗传算法在平均预测误差最小化准则下求解模型的最优参数。最后分别应用传统GM(1,1)模型和FourierGM(1,1)模型预测森林火灾扑火经费,结果表明:本文提出的新方法能够较好地描述时间序列中的周期性振荡特征,其预测精度显著地高于传统GM(1,1)模型。
In view of the characteristics of small sample and oscillation in time series of demand of disaster emergency supplies,this paper puts forward Fourier-GM （1,1） model for forecasting demand of emergency supplies. Firstly,GM （1,1） model with adaptive background value is established for a given the small sample oscillating sequences, then Fourier series is applied to describe the periodic oscillation information contained in residuals,thus new forecasting model Fourier-GM （1,1） is constructed. On this basis, a genetic algorithms is used to solve the optimal parameters under the criterion of average relative errors minimization. Finally, the proposed Fourier-GM （1, 1） model and the traditional GM （1, 1） model are employed to forecast suppression funds for forest fire. The results show that the new method proposea in this paper can better describe the periodic oscillation characteristics of time series, and that the prediction precision is higher than that of the traditional GM （1,1） model.
出处
《系统工程》
CSSCI
CSCD
北大核心
2013年第8期60-64,共5页
Systems Engineering
基金
国家自然科学基金资助项目(71101132)
国家社会科学基金资助项目(12BGL064)
