交通运输系统工程与信息

交通运输系统工程与信息 ›› 2021, Vol. 21 ›› Issue (6): 84-95.DOI: 10.16097/j.cnki.1009-6744.2021.06.010

• 系统工程理论与方法 • 上一篇    下一篇

基于乘客感知的多模式公交服务质量差异性研究

朱兴林*，姚亮，刘泓君，叶拉森∙库肯，克然木∙司马义   

  1. 新疆农业大学，交通与物流工程学院，乌鲁木齐 830052
  • 收稿日期:2021-07-13 修回日期:2021-09-01 接受日期:2021-09-02 出版日期:2021-12-25 发布日期:2021-12-23
  • 作者简介:朱兴林(1971- )，女，新疆乌鲁木齐人，副教授。
  • 基金资助:
    国家自然科学基金；新疆维吾尔自治区研究生创新项目；新疆维吾尔自治区研究生教育教 学改革项目

Investigating Differences in Service Quality of Multi-mode Public Transit Based on Passenger Perception

ZHU Xing-lin* , YAO Liang, LIU Hong-jun, ERASEL·Kuken, KERAM·Iesmayil   

  1. College of Transportation and Logistics Engineering, Xinjiang Agricultural University, Urumqi 830052, China
  • Received:2021-07-13 Revised:2021-09-01 Accepted:2021-09-02 Online:2021-12-25 Published:2021-12-23
  • Supported by:
    National Natural Science Foundation of China(2018D01A21)；Postgraduate Innovation Project of Xinjiang Uygur Autonomous Region, China(XJ2021G165)；Postgraduate Education and Teaching Reform Project in Xinjiang Uygur Autonomous Region, China(XJ2020GY15)

PDF

摘要： 针对多模式公交服务质量的差异性，提出一种以TAN贝叶斯网络与支持向量机(SVM)为 基础的多方法组合评价模型，评价各模式公交服务质量，并预测检验了指标优化效应。首先，采 集2015—2018年市区线路、郊区线路、快速公交、定制公交的服务质量调查数据，基于公交方式截 面，使用TAN贝叶斯网络推理得出各因素对乘客满意度的影响能力及潜在影响关系；其次，结合 IPA分析方法，基于时间截面定位分析各模式公交选取指标的服务水平，辨识各模式公交需优化 的主要指标；最后，选取多层感知器(MLP)、长短时记忆神经网络(LSTM)与SVM进行对比，验证 了SVM回归预测的精准度，并采用SVM与OAT方法预测得到各模式公交满意度变化及指标敏 感性，参照因素间潜在的影响关系提出优化方案。结果表明：4种公交模式的服务质量影响关系 网络具有差异性，各模式公交均存在车厢拥挤问题，市区线路的指标优化效果最佳，正效应为 36.4%；市区和郊区线路均应与其他模式公交匹配发车计划，通过缩短乘客候车时长可分别提升 39%、32.2%的乘客满意度；快速公交需提升车辆行驶的稳定性，定制公交需要调整线路规划，减 少乘客乘车总时长，优化为整体服务质量提升带来的正效应分别为42.7%和37.4%。

关键词: 城市交通, 服务质量, TAN贝叶斯网络, 多模式公交, 支持向量机, 重要性绩效分析

Abstract: To handle the differences in the service quality of multi-mode public transit, an evaluation model based on TAN Bayesian network and Support Vector Machine was proposed to evaluate the service quality of each mode of public transit, and the index optimization effect was predicted and tested for the differences in multi-mode bus service quality. The service quality survey data from 2015 to 2018 were collected from urban routes, suburban routes, BRT, and customized buses, and the ability of each factor to influence passenger satisfaction and their potential influence relationship were derived by the TAN Bayesian network inference based on a cross-section of transit modes. Combined with the IPA method, the service level of public transit index is positioned based on the time section, and the main indicators that need to be optimized are extracted. Finally, the Multi-Layer Perceptron, Long-Short-Term Memory neural network, and SVM were taken as alternatives, and the accuracy of the SVM regression prediction was verified. The SVM and OAT methods were used to predict the changes in satisfaction and indicator sensitivity of each mode of public transportation, and the optimization scheme was proposed with reference to the potential influence relationship between factors. The results show that there are differences in the network of service quality influence relationships among the four modes of public transportation, and all modes of public transportation have the problem of cabin crowding, with the best indicator optimization effect of 36.4% for urban routes; both urban and suburban routes should match the departure schedule with other modes of public transportation, and can improve passenger satisfaction byshortening passenger waiting time by 39% and 32.2%, respectively; BRT needs to improve vehicle Bus Rapid Transit needs to improve vehicle stability, while customized buses need to adjust route planning to reduce the passenger travel time, and the positive effect of optimization on overall service quality improvement is 42.7% and 37.4%, respectively.

Key words: urban traffic, service quality, TAN Bayesian network, multi-mode public transport, support vector machine, importance performance analysis

中图分类号: