基于行程时间的公交运行可靠度评价及影响因素分析  

Evaluation of Bus Operation Reliability and Analysis of Influencing Factors Based on Travel Time

作　　者：翁剑成[1] 赵世昌 林鹏飞 孔宁 钱慧敏 Weng Jiancheng;ZHAO Shichang;LIN Pengfei;KONG Ning;QIAN Huimin(Beijing University of Technology,Beijing Key Laboratory of Traffic Engineering,Beijing 100124,China;Beijing University of Technology,College of Computer Science,Beijing 100124,China;CCCC Intelligence Transportation Company Limited,Tianjin 300210,China;Beijing Municipal Transportation Operations Coordination Center,Beijing 100161,China)

机构地区：[1]北京工业大学交通工程北京市重点实验室,北京100124 [2]北京工业大学计算机学院,北京100124 [3]中交智运有限公司,天津300210 [4]北京市交通运行监测调度中心,北京100161

出　　处：《交通信息与安全》2024年第6期163-171,共9页Journal of Transport Information and Safety

基　　金：国家自然科学基金项目(52302381、52072011);北京市教育委员会科学研究计划项目(KM202310005025)资助。

摘　　要：公交运行过程中会受到多种内外部因素干扰,为精准评价公交运行可靠度并量化分析各影响因素。基于公交到站时间数据计算区间行程时间,通过动态阈值和变异系数以及归一化处理,研究了1种可反映不合理延误影响和区间行程时间波动性的公交运行可靠度评价方法,实现不同线路、不同时段公交运行可靠度的横、纵向对比,解决了基于时刻表偏差的公交运行可靠度评价方法不适用于高频服务公交线路的问题。为克服现有研究影响因素考虑单一、以定性分析为主的局限,从站点客流、公交线站属性、道路条件等维度构建8种公交运行可靠度影响因素,利用随机森林模型构建可靠度影响模型,并与支持向量机与反向传播神经网络(back propagation,BP)模型进行精度对比,结合相对重要度和部分依赖图量化分析各影响因素。选取北京市2019年1月9条公交线路的多源公交数据进行实证分析。结果表明:研究提出的评价方法具有良好的有效性,可精准识别早晚高峰期间公交运行的不可靠。采用随机森林构建影响模型的精度最高,相较于支持向量机与BP神经网络分别提升20.38%和49.88%。模型揭示了各个因素的非线性影响机理并确定了有效阈值区间,站点间距、公交区间速度与公交专用道占比是影响公交运行可靠度的关键因素,相对重要度依次为26.9%、25.1%和24.1%。此外,当站点间距在600~800 m之间时,可靠度相较于250 m提升约12.5%;可靠度与公交区间速度呈正相关关系,最高可提升约7%;当公交专用道占比达到60%以上时可靠度显著提升,当占比达到95%时,可靠度提升约6.5%;当途径路段的交叉口数量从1个增加至3个时,可靠度下降约4%;为保证良好的可靠度,公交站台服务的公交线路数不宜超过3条。Bus operation is subject to various internal and external factors.To accurately evaluate bus operation reli-ability and quantitatively analyze the influencing factors.this study calculated the interval travel time based on bus arrival time data.It established a bus operation reliability evaluation method that can reflect the impact of unreason-able delays and the variability of interval travel time by calculating the dynamic threshold probability and coeffi-cient of variation and normalization processing.This method achieves horizontal and vertical comparison of bus op-eration reliability for different routes and different time periods,solving the problem that the bus operation reliabili-ty evaluation method based on schedule deviation is not applicable to high-frequency service bus routes.To address the limitations of existing research,which primarily focuses on single-factor considerations and qualitative analy-sis,eight influencing factors of bus operation reliability are constructed from perspectives such as station passenger flow,bus route and stop attributes,and road conditions.A Random Forest model is utilized to develop an impact model for bus operation reliability,and its accuracy is compared with that of support vector machine(SVM)and back propagation(BP)Neural Network model.This study used relative importance analysis with partial dependence plots to quantitatively identify key factors and reveal the impact mechanisms.The study uses multi-source bus data from 9 bus routes in Beijing from January 2019 for empirical analysis.The results show that the proposed evalua-tion method is effective in accurately identifying unreliable bus operations during morning and evening peak hours.The accuracy of the impact model constructed using random forest(RF)is the highest,with improvements of 20.38%and 49.88%compared to SVM and BP Neural Networks,respectively.Key factors influencing reliability in-clude bus stop spacing,bus section speed,and the proportion of dedicated bus lanes,with relative importance values of 26

关 键 词：公交运行可靠度 区间行程时间 随机森林模型 部分依赖图 

分 类 号：U491.17[交通运输工程—交通运输规划与管理]