城市轨道交通服务质量评价是行业监管部门为监督和评估轨道交通运营服务水平，定期开展的考核工作。传统的城市轨道交通服务质量评价采用比较各线路总体得分的形式，对乘客的主体感知和真实需求的探讨较少。为提高城市轨道交通服务质量，基于乘客的感知及需求构建城市轨道交通服务质量评价指标体系，建立基于SEM及贝叶斯网络的综合评价模型。通过SEM评估模型中变量之间的稳健关系，并利用贝叶斯推理与诊断对其进行辅助决策，可以更准确地揭示各变量对服务质量的影响程度，达到有针对性的优化服务方案，多维度的满足乘客需求。最后以昆明市轨道交通为例，结果表明：设施可靠、人工服务与出行安全对服务质量的影响幅度较大（路径系数分别为0.50、0.39、0.38；变化幅度分别为2.86%、2.45%、2.20%）。同时本文所提出模型的偏差信息准则 ，验证了该模型的可行性与适用性，以期为地铁运营者如何提高服务质量，吸引更多客流提供借鉴。

The evaluation of urban rail transit service quality is a regular assessment work carried out by the industry supervisory department to monitor and evaluate the operation service level of rail transit. The traditional evaluation of urban rail transit service quality takes the form of comparing the overall scores of each line, with less exploration of the main perceptions and real needs of passengers. To improve the service quality of urban rail transit, an evaluation index system of urban rail transit service quality is constructed based on passengers' perceptions and demands and developed a comprehensive evaluation model based on SEM and Bayesian network. By evaluating the robust relationships among variables in the model through SEM and using Bayesian inference and diagnosis to assist in their decision making, the degree of influence of each variable on service quality can be more accurately revealed to achieve targeted optimization of service solutions and meet the multidimensional needs of passengers. Finally, take Kunming rail transit as an example. The results show that the magnitude of the influence of facility reliability, manual service, and travel safety on service quality is large (the path coefficients are 0.50, 0.39, and 0.38; the variation is 2.86%, 2.45%, and 2.20%, respectively). Meanwhile, the Deviance Information Criterion of the model proposed in this paper verifies the feasibility and applicability of the model. It can provide a reference for metro operators on how to improve service quality and attract more passenger flow.

U231???????????????

云南省科技厅科技计划面上项目(2019FB086)；昆明理工大学引进人才科研启动基金资助项目(KKSY201802016)