武广高速铁路列车晚点恢复时间预测的随机森林模型  

Random Forest Prediction Model for Wuhan-Guangzhou HSR Primary Train Delays Recovery

作　　者：黄平 彭其渊[1,2] 文超 杨宇翔[1,4] HUANG Ping;PENG Qiyuan;WEN Chao;YANG Yuxiang(School of Transportation and Logistics,Southwest Jiaotong University,Chengdu 410031,China;National United Engineering Laboratory of Integrated and Intelligent Transportation,Chengdu 610031,China;Railway Research Center,University of Waterloo,Waterloo N2L3G4,Canada;Institute of Transport Science,RWTH Aachen University,Aachen 52074,Germany)

机构地区：[1]西南交通大学交通运输与物流学院,四川成都610031 [2]综合交通运输智能化国家地方联合工程实验室,四川成都610031 [3]滑铁卢大学铁路研究中心,滑铁卢N2L3G1 [4]亚琛工业大学交通科学研究所,亚琛52074

出　　处：《铁道学报》2018年第7期1-9,共9页Journal of the China Railway Society

基　　金：国家重点研发计划(2017YFB1200701);国家自然科学基金(61503311);中国铁路总公司科技开发计划项目(2016X008-J)

摘　　要：基于列车运行实绩的列车晚点恢复模型是铁路晚点管理的重要内容,是运行图优化和行车指挥的理论基础和依据。为了研究高速列车初始晚点恢复的机理,进行初始晚点恢复预测,本文以武广高速铁路列车运行实绩数据为研究基础,将列车在初始晚点站的晚点时间(PD)、列车晚点后经停各站的总停站缓冲时间(TD)、列车晚点后经停各区间的总区间缓冲时间(RB),以及标识列车是否晚点通过株洲西—长沙南区间的0-1变量(ZC)作为自变量,运用R语言编程建立了以晚点恢复时间(RT)为因变量的高速列车初始晚点恢复随机森林回归模型。对275个测试集数据的预测结果表明:模型允许误差在3min情况下,模型的预测精度能达到90%以上。随机森林模型与多元线性回归模型、支持向量机模型的对比表明,随机森林模型具有最优的预测精度。Delay recovery modelling based on real-world train operation records is an essential issue to the delay management,as well as the fundamental theory and principle of timetable rescheduling and railway traffic control.In this paper,to study the primary delay recovery prediction,based on the real-world train operation records from of Wuhan-Guangzhou high-speed railway,four factors,named the primary delay（PD）,subsequent total dwell buffer time at the dwell stations of the delayed train（TD）,subsequent total running buffer time in passing through sections by the delayed train（RB）and a 0-1 discrete variable ZCthat distinguishes whether a train passes through the section of ZhuzhouXi-ChangshaNan in delayed status were selected as the independent parameters for modelling.A random forest regression prediction model,where the delay recovery time（RT）is the dependent variable was established.The test results from 275 testing datasets show that the prediction precision of the model can reach more than 90% with apermissible error of 3 min.The prediction precision comparison of the random forest model with the multiple linear regression model and the support vector machine model shows that the random forest model has the best performance.

关 键 词：高速铁路 初始晚点 列车运行实绩 晚点恢复 随机森林模型 

分 类 号：U292.1[交通运输工程—交通运输规划与管理]