基于SARIMA-SVR模型的铁路货运量预测方法  

Railway Freight Volume Prediction Based on SARIMA-SVR Model

作　　者：钱名军[1,2] 李明鲡 黄鑫 QIAN Mingjun;LI Mingli;HUANG Xin(School of Traffic and Transportation,Lanzhou Jiaotong University,Lanzhou 730070,Gansu,China;Key Laboratory of Railway Industry on Plateau Railway Transportation Intelligent Management and Control,Lanzhou Jiaotong University,Lanzhou 730070,Gansu,China)

机构地区：[1]兰州交通大学交通运输学院,甘肃兰州730070 [2]兰州交通大学高原铁路运输智慧管控铁路行业重点实验室,甘肃兰州730070

出　　处：《铁道运输与经济》2024年第9期83-94,共12页Railway Transport and Economy

基　　金：兰州交通大学青年基金项目(2014029);甘肃省教育厅高等学校创新基金项目(2020A-038);甘肃省教育厅双一流重大科研项目(GSSYLXM-04)。

摘　　要：鉴于铁路货运量受多种外部因素影响呈现显著的随机波动特征而难以准确预测,提出了SARIMA-SVR预测模型。首先,对全国铁路月度货运量序列进行季节时间序列(SARIMA)建模,得到模型的初始预测值及预测残差。其次,构建支持向量机(SVR)回归预测模型,将影响铁路货运量的外部因素作为模型输入项,SARIMA模型预测残差序列、月度货运量序列分别作为模型输出项,由此分别获得SARIMA模型预测残差的优化值以及SVR模型的货运量预测值。三是将优化后的SARIMA模型预测残差与其初始预测值相加,得到优化后的SARIMA模型预测值。四是再对优化后的SARIMA模型预测值和SVR模型预测值进行加权求和,得到SARIMA-SVR模型的预测结果。最后,对SARIMA-SVR模型进行消融实验验证模型有效性,并将该模型与经典预测模型进行测算精度对比。结果表明,SARIMA-SVR模型的预测精度优于单一模型和经典预测模型,在货运量预测方面具有良好的适用性。As the railway freight volume is hard to predict due to its significantly random fluctuation caused by a variety of external factors,the SARIMA-SVR prediction model was proposed.First of all,seasonal time series(SARIMA)modeling was carried out on the monthly freight volume series of national railways to obtain the initial predicted value and prediction residuals.Then,the support vector regression machine model was constructed.After taking the external factors affecting railway freight volume as model inputs and prediction residuals series of the SARIMA model as well as monthly freight volume series as model outputs,optimized values of prediction residuals of the SARIMA model and freight volume predicted values of the SVR model were obtained,respectively.Next,the optimized SARIMA model prediction residuals were added with the initial prediction values to obtain the optimized SARIMA model predicted values.After that,the optimized SARIMA model predicted values and SVR model predicted values were weighted and summed to obtain the SARIMA-SVR model predicted results.Finally,ablation experiments were conducted on the SARIMA-SVR model to verify the model’s validity,and the model was also compared with the classical model in terms of measured accuracy.The results show that the prediction accuracy of the SARIMA-SVR model is better than that of the single model and the classical prediction model,and can be well applied to predict freight volume.

关 键 词：铁路运输 货运量预测 SARIMA-SVR模型 季节性时间序列 支持向量机 

分 类 号：U294.13[交通运输工程—交通运输规划与管理]