基于径向基神经网络的铁路货运量预测

Study on Prediction of Railway Freight Volumes Based on RBF Neural Network

货运量预测是铁路运输部门一项重要工作，因此，关于铁路货运量预测理论和方法的研究一直是一个热点。但是，铁路货运量受多种因素影响，且各因素的作用机制通常不能或无法用精确的数学语言来准确描述。本文采用径向基函数（RBF）神经网络对货运量进行分析及预测。通过对1989～2002年全国铁路货运量的历史数据分析处理后，得到铁路货运量增长量的时间序列，将时间序列视为一个从输入到输出的非线性映射，引入RBF神经网络来进行非线性映射的逼近。对网络进行学习与训练仿真实验后，用2003～2004年的增长量进行模型检验，并与BP神经网络、灰色预测模型预测结果进行对比，结果表明，应用RBF神经网络对铁路货运量进行短期预测预测精度更高、效果更好。

Railway freight volume forecasting is an important responsibity of the railway transportation management departments, so the theory and method of railway freight volume forecasting remain a focus in research all the time. Railway freight volumes are influenced by multiple factors, and the action mechanisms of these factors are usually unable to be described with accurate mathematical linguistic forms. The radial basis function （RBF） neural network is adopted to analyze and predict freight volumes. Through processing the statistic data of freight volumes from 1989 to 2002, we get the temporal sequence of increments of freight volumes, which can be regarded as mapping of nonlinear approximations from input to output. The RBF neural network is introduced to solve the nonlinear approximation issue. After the network learning and training simulation experiments are completed, the model is established and verified by use of freight volume increments from 2003 to 2004. Comparing with the BP neural network model and Gray Prediction Method GM（1,1） , the RBF neural network is more precise and effective in railway freight volume short-term forecast.