基于神经网络与Holt-Winters模型的铁路货运量组合预测  

Combination Forecast of Railway Freight Volume Based on Holt-Winter Model and Neural Network

作　　者：王庆荣[1] 

机构地区：[1]兰州交通大学电子与信息工程学院,甘肃兰州730070

出　　处：《兰州交通大学学报》2010年第4期122-125,共4页Journal of Lanzhou Jiaotong University

基　　金：教育部"春晖计划"项目(Z2005-1-62008);甘肃省自然科学基金(ZS022-A25-036)

摘　　要：铁路货运量的预测结果直接影响到铁路运输计划以及有关运输政策的制定.在有关文献研究的基础上,提出了基于Holt-Winters模型和神经网络的铁路货运量组合预测方法,并为Holt-Winters模型和神经网络模型设计了相应的算法,同时,采用最小绝对值法,确定了组合模型中各单项预测模型的权重.最后利用实例验证了该组合预测方法的有效性和实用性.预测结果分析表明:最大预测误差为9.13%,最小预测误差为2.38%,说明本文提出的组合预测模型充分考虑了多指标的共同作用,具有一定的科学性.Forecasting result of railway freight volume influences directly on railway plan and enacting of transportation policy.Based on related document,a combination forecasting model of Holt-Winter model and neural network for railway freight volume is set up in this paper,and a corresponding algorithm is designed,then the weight of individual forecasting model is definited with the least Absolute value method.Finally,the availability and practicability of this model are proved.The result of which shows that the maximum predictive encoding is 9.13% and the mimimum predictive encoding is 2.38%,and also shows that this model is more scientific.

关 键 词：铁路 货运量预测 Holt-Winters模型 神经网络 

分 类 号：U291.1[交通运输工程—交通运输规划与管理]