摘要
铁路货运量与其影响因素之间存在着复杂的非线性关系,传统的BP神经网络模型能对非线性系统进行很好的拟合,但模型的预测能力不强.通过单位根检验,可知铁路货运量及其影响因素的时序列数据是非平稳的.本文通过分析BP神经网络的传递函数对非平稳时间序列预测的不利影响,提出用差分法对输入数据进行预处理,在此基础上建立了铁路货运量预测的改进BP神经网络模型,并通过实例计算说明了这种改进BP神经网络方法对提高铁路货运量预测精度的有效性,最后利用该模型对2006—2010年的铁路货运量进行了预测.
The relation between railway freight volume and its influence factors is complex and nonlinear. The BP Neural Network can simulate the nonlinear system perfectly, but its forecast ability is deficient. The umsteadiness of the time series of railway freight volume and its influence factors are proved by unit root test. With the analysis of the transfer function, an improved BP Neural Network model is built on the basis of using the difference method to deal with the input data. At last, the effectiveness of the new model in railway freight volume forecast is proved through an experimental computation, and then the railway freight volume in 2006 -2010 years is forecast by the model.
出处
《交通运输系统工程与信息》
EI
CSCD
2006年第6期158-162,共5页
Journal of Transportation Systems Engineering and Information Technology
关键词
BP神经网络
铁路货运量
单位根检验
预测
BP neural network
railway freight volume
unit root test
forecast
