基于EMD-APSO-SVR组合模型的铁路货运量预测  

Rail Freight Volume Forecasting Based on Combined EMD-APSO-SVR Model

作　　者：韩纯良 李默涵 洒雨 周琳 吴林鸿 薛锋[2,3,4] HAN Chunliang;LI Mohan;SA Yu;ZHOU Lin;WU Linhong;XUE Feng(Market Monitoring and Evaluation Center,National Railway Administration of China,Beijing 100070;School of Traffic and Logistics,Southwest Jiaotong University,Chengdu 611756,China;National Engineering Laboratory of Integrated Transportation Big Data Application Technology,Southwest Jiaotong University,Chengdu 611756,China;National United Engineering Laboratory of Integrated and Intelligent Transportation,Southwest Jiaotong University,Chengdu 611756,China)

机构地区：[1]国家铁路局市场监测评价中心,北京100070 [2]西南交通大学交通运输与物流学院,成都611756 [3]西南交通大学综合交通大数据应用技术国家工程实验室,成都611756 [4]西南交通大学综合交通运输智能化国家地方联合工程实验室,成都611756

出　　处：《综合运输》2024年第4期132-140,180,共10页China Transportation Review

基　　金：国家铁路局科研项目(市场委合2022-3号)。

摘　　要：为了提高铁路货运量预测的准确性,减少类似新冠疫情这样突发事件对现有预测结果的不稳定影响,提出一种基于EMD和APSO优化的SVR组合预测模型。在综合考虑铁路货运量受GDP、铁路营业里程、原煤产量、钢材产量、第二产业增加值以及新冠疫情等因素情况下,对选取的货运量序列进行EMD处理,得到不同时间尺度下的IMF和残差;通过APSO优化SVR模型的参数,并利用优化后的组合模型对各IMF分量分别进行预测,相加得到最终结果。将APSO-SVR模型与EMDAPSO-SVR模型的预测状况进行对比,结果表明,本文建立的EMD-APSO-SVR模型的预测结果误差更小,其预测值与真实值的测定系数高于APSO-SVR模型;前者的平均绝对百分比误差仅有0.22%,能有效提升铁路货运量短期内的预测精度。In order to improve the accuracy of rail freight forecasting and reduce the unstable impact of unexpected events similar to COVID-19 on the existing forecasting results,a combined prediction model for SVR based on EMD and APSO optimization is proposed.Under the comprehensive consideration of railroad freight volume subject to GDP,railroad business mileage,raw coal production,steel production,secondary industry added value and COVID-19,the selected freight volume series are processed by EMD to obtain IMFs and residuals under different time scales;the parameters of the SVR model are optimized by the APSO algorithm under an adaptive variation strategy;and this combined model is used to predict each IMF component separately and sum up to obtain the final results.The prediction status of the APSO-SVR model is compared with that of the EMD-APSO-SVR model,and the results show that the prediction result error of the EMD-APSO-SVR model established in this paper is smaller and the coefficient of determination of its predicted and true values is higher than that of the APSO-SVR model;the average absolute percentage error of the former is only 0.22%,which can effectively improve the short-term railroad freight volume prediction accuracy.

关 键 词：铁路货运量预测 新冠疫情 经验模态分解 自适应粒子群算法 支持向量机回归模型 

分 类 号：U2-9[交通运输工程—道路与铁道工程]