RBF神经网络在铁路货运量预测中的应用

基于回归和时间序列模型的传统预测方法以及目前较为常用的灰色预测和BP神经网络预测方法，建立了RBF神经网络模型对全国铁路货运量进行详细分析和预测。利用铁路货运量的原始数据构造时间序列，并对时间序列进行分析和相应的处理。将处理后的数据构造为一个非线性映射，利用RBF神经网络进行逼近。利用Matlab对灰色预测、BP神经网络预测和RBF神经网络预测模型进行仿真实验，得出3种预测模型的平均相对误差，分别为7.67%、4.79%和1.31%。表明RBF神经网络预测方法的预测精度比另外两种预测方法高很多，可为铁路货运量预测研究提供方法支撑。

RBF neural network applied to prediction of railway freight volumes

Based on the prediction methods of regression model, times series model, as well as the commonly used prediction methods of grey model and BP neural network, RBF neural network model was established in this case to analysis and predict the national railway freight volume in detail. The original data of railway freight volume was used to construct times series, analyze and deal with the times series. The processed data were constructed as a nonlinear mapping. RBF neural network was used to approach it. The grey model, BP neural network model and RBF neural network model were simulated by using Matlab. It was concluded that average relative errors of the three kinds of prediction model were 7.67%, 4.79% and 1.31% respectively. The results showed that the prediction method based on RBF neural network was better than other, could provide support for railway freight volume forecasting.