基于CV-GRNN的铁路货运量预测

Railway Freight Volume Forecasting Based on CV-GRNN

精确的铁路货运量预测对铁路建设规划以及企业决策运营是一项非常重要的手段。由于铁路货运市场供需不平衡,使得铁路货运量预测具有复杂性和非线性。广义回归神经网络(Generalized Regression Neural Network,GRNN)弥补了很多传统建模方法的不足,在多因素、非线性映射上具有良好的性能,通常采用试算法确定GRNN模型参数spread值,计算较为繁琐,寻优效率较低。以1990~2013年铁路货运量数据为例进行研究,对广义回归神经网络预测方法进行改进,引入交叉验证算法来寻优广义回归神经网络的spread值,然后将最优spread值赋予广义回归神经网络进行铁路货运量预测。仿真结果表明,改进后的广义回归神经网络,预测误差得到降低,预测精度优于基本广义回归神经网络算法,提出预测模型是可行且有效的。

Accurate prediction of railway freight volume is a very important means for railway construction planning and enterprise decision-making operation.Due to the imbalance of supply and demand in the railway freight market,the prediction of railway freight volume becomes more complex and nonlinear.The generalized regression neural network(GRNN)makes up for the shortcomings of many traditional modeling methods,and has good performance in multi factor,nonlinear mapping.However,it is usual to use the trial-and-error method to determine the optimal spread value,so the calculation is cumbersome and inefficient.In this paper,the cross validation algorithm(CV)is introduced to optimize the spread value of GRNN(CV-GRNN),and then the optimal spread value is given to GRNN for forecasting the railway freight volume.