基于集成学习的公交车辆到站时间预测模型研究  

Research on Bus Arrival Time Prediction Model Based on Ensemble Learning

作　　者：荆灵玲[1] 解超[2,3] 王安琪 JING Lingling;XIE Chao;WANG Anqi(AVIC Institute of Geotechnical Engineering Co.,Ltd,Beijing 100098,China;China Transport Telecommunications&Information Center,Beijing 100011,China;ZJX Technologies Inc.,Beijing 100007,China;North China University of Technology,Beijing 100144,China)

机构地区：[1]中航勘察设计研究院有限公司,北京100098 [2]中国交通通信信息中心,北京100011 [3]中交信有限责任公司,北京100007 [4]北方工业大学,北京100144

出　　处：《重庆理工大学学报（自然科学）》2019年第10期47-53,共7页Journal of Chongqing University of Technology：Natural Science

基　　金：贵州省交通厅科技项目“城市道路交通大数据智能监管云平台”(GZWH-2017-1416)

摘　　要：公交车是城市交通系统的主要组成部分,准确预测公交车辆到站时间作为城市智能交通系统(ITS)的重要应用,可以有效提升公交车乘客的乘坐体验,增加公交出行的吸引力。为此,提出一种基于集成学习方法的公交车到站时间预测模型,利用联合多种弱模型的方法,确定优化目标,将公交车到站时间相关的影响因素进行特征化后,基于海量历史数据训练机器学习模型,进而预测公交到站时间。实例分析和验证结果表明,GBDT方法的预测性能明显优于其他方法,可显著提高公交到站时间预测的准确性。Bus is the main component of urban traffic system.As an important application of urban intelligent transportation system(ITS),accurate prediction of bus arrival time can effectively enhance the passenger experience and increase the attraction of bus travel.In this paper,a bus arrival time prediction model based on ensemble learning method is proposed.By combining several weak models,the optimization objective is determined.After characterizing the factors related to bus arrival time,the machine learning model is trained based on massive historical data,and then the bus arrival time is predicted.On this basis,the paper proves that the GBDT method is superior to other methods in predicting the arrival time of public transport through case analysis and verification,and it can significantly improve the accuracy of bus arrival time prediction.

关 键 词：历史路况 实时路况 特征提取与计算 集成学习 GBDT 

分 类 号：U491.17[交通运输工程—交通运输规划与管理]