祝玥宁,上官伟,柴琳果,蔡伯根,王伟杰

• 1:北京交通大学自动化与智能学院
• 2:北京交通大学轨道交通控制与国家重点实验室
• 3:北京交通大学自主智能与无人系统研究中心

摘要(Abstract)：

为提升城市轨道交通应对突发事件的风险预测能力，基于北京地铁近年来运营突发事件的真实数据，引入及时安全管理理念，提出基于可解释集成学习的城轨列车运行风险影响预测性评估方法。首先，根据历史真实数据，探究影响轨道交通列车运行的风险致因及其影响范围，并结合风险导致突发事件概率及影响后果严重程度，利用动态风险权重以风险评分为依据评估风险影响严重等级。其次，构建Stacking集成学习模型，利用SMOTE过采样改善样本不均衡问题，并结合Optuna寻找基模型与元模型的的最优参数组合。结果表明：基于Stacking集成学习模型，对突发事件风险影响严重等级的预测分类准确率可达97.12%,5项指标均较单一机器学习基模型性能均有提高。最后，引入了SHAP可解释性分析，通过计算各风险影响特征对模型预测结果的边际贡献度，获得特征重要性排序与关键风险致因，旨在确保防控手段重点关注由Stacking集成学习模型经可解释分析识别出的、具有统计显著性的风险致因，进而实现预测精度与决策针对性的同步提升，为早期干预、预警发布、主动风险防控策略的实施提供了依据。

关键词(KeyWords)： 城市轨道交通;风险致因分析;风险预测;可解释性分析

Abstract:

Keywords:

基金项目(Foundation): 北京市自然科学基金(L231003);; 国家自然科学基金(52272328);; 北京高等学校卓越青年科学家(JWZQ20240101010);; 中央高校基本科研业务费专项资金资助(2025YJS111)

作者(Author): 祝玥宁,上官伟,柴琳果,蔡伯根,王伟杰

参考文献(References)：

• [1] Li Y,Ding S,Wang S,et al.Safety assessment method of high-speed rail interval structure based on weighted bayesian network[J].KSCE Journal of Civil Engineering,2024,28(8):3286-3300.
• [2] 宋晗炜.基于设备状态信息的FAO系统运行风险态势评估[D].北京：北京交通大学，2021.Song Hanwei.Risk situation assessment of FAO system operation based on equipment status[D].Beijing:Beijing Jiaotong University,2021.
• [3] 潘恒彦，王永岗，李德林，等.基于交通冲突的长纵坡路段追尾风险评估及预测[J].吉林大学学报，2023,53(5):1355-1363.Pan Hengyan,Wang Yonggang,Li Delin,et al.Evaluating and forecasting rear-end collision risk of long longitudinal gradient roadway via traffic conflict[J].Jilin Daxue Xuebao,2023,53(5):1355-1363.
• [4] He J,Fan X,Wu L.Identification and Evaluation of urban rail transit operation risk factors based on entropy-AHP hybrid constrained DEA method[J].Journal of Advanced Transportation,2022:6241096.
• [5] 王宗耀，上官伟，彭聪，等.基于CODAS的CBTC车载系统子系统故障风险评估与维修优先级决策方法[J].铁道学报，2023,45(8)87-98.Wang Zongyao,Shangguan Wei,Peng Cong,et al.Risk assessment and maintenance priority decisionmaking method for failure of subsystem of CODAS-based CBTC onboard system[J].Journal of the China Railway Society,2023,45(8):87-98.
• [6] Ju W,Wu J,Kang Q,et al.Fire risk assessment of subway stations based on combination weighting of game theory and topsis method[J].Sustainability,2022,14(12):7275.
• [7] Kuang X,Sun S,Zeng Q.Risk assessment of information security in communication-based train control systems based on ahp fuzzy comprehensive evaluation[C]// 7th International Conference on Artificial Intelligence and Big Data (ICAIBD).Piscataway:IEEE,2024:247-252.
• [8] Mbaye S,Jones G,Davies M D.Process Flow modeling for an in-time aviation safety management system[J].INCOSE International Symposium,2022,32(1):888-902.
• [9] Ellis K K,Vincent M J,Prinzel L J,et al.The in-time aviation safety management system (IASMS) concept of operations for evolving safety in air traffic management (ATM) Services[C]//AIAA Aviation Forum and Ascend.Reston:AIAA,2022:1-10.
• [10] Ahmed G,Sheltami T,Deriche M,et al.An energy efficient IoD static and dynamic collision avoidance approach based on gradient optimization[J].Ad Hoc Networks,2021,118:102519.
• [11] 张俊英，王勇.通信时延和感知时延对车辆编队稳定性的影响[J].中国公路学报，2023,36(4):202-220.Zhang Junying,Wang Yong.Effects of communication and sensing delays on stability of vehicular platoons[J].China Journal of Highway and Transport,2023,36(4):202-220.
• [12] Li L,Gan J,Qu X,et al.A novel graph and safety potential field theory-based vehicle platoon formation and optimization method[J].Applied Sciences,2021,11(3):958.
• [13] Zhu Y,Shangguan W,Chen J,et al.Risk predictive deduction and propagation chain mining for URT emergencies based on interpretable ensemble learning:a new approach to in-time safety management[J].Reliability Engineering & System Safety,2026,272:112490.
• [14] 赵海涛，程慧玲，丁仪，等.基于深度学习的车联边缘网络交通事故风险预测算法研究[J].电子与信息学报，2020,42(1):50-57.Zhao Haitao,Cheng Huiling,Ding Yi,et al.Research on traffic accident risk prediction algorithm of edgeInternet of vehicles based on deep learning[J].Journal of Electronics and Information Technology,2020,42(1):50-57.
• [15] Liu J,Zhang Y,Han J,et al.Intelligent hazard-risk prediction model for train control systems[J].IEEE Transactions on Intelligent Transportation System,2020,21(11):4693-4704.
• [16] Keramati A,Lu P,Moatari-Kazerouni A,et al.Evaluating crash severity at highway-rail grade crossings using an analytic hierarchy process-based hazard index model[J].Accident Analysis and Prevention,2025,211:107918.
• [17] Gao W,Lu Y,Wang N,et al.Measurement and prediction of subway resilience under rainfall events:An environment perspective[J].Transportation Research Part D-Transport and Environment,2024,136:104479.
• [18] Baptista M,Goebel K,Henriques E.Relation between prognostics predictor evaluation metrics and local interpretability SHAP values[J].Artificial Intelligence,2022,306:103667.
• [19] Ma X,Huo Z,Lu J,et al.Deep forest with SHapley additive explanations on detailed risky driving behavior data for freeway crash risk prediction[J].Engineering Applications of Artificial Intelligence,2025,141:109787.
• [20] 邓院昌，蒋昀轩，陶胜芹.基于可解释集成学习的异常驾驶行为风险识别方法[J].交通运输系统工程与信息，2025,25(2):180-189.Deng Yuanchang,JIANG Junxuan,TAO Shengqin.Risk identification method for abnormal driving behavior based on interpretable ensemble learning model[J].Journal of Transportation Systems Engineering and Information Technology,2025,25(2):180-189.
• [21] Gao T,Chen J,Xu H.Data-driven train delay prediction incorporating dispatching commands:an XGBoost metaheuristic framework[J].IET Intelligent Transport Systems,2024,18(10):1777-1796.
• [22] 申中帅，鲍学英，刘北胜，等.铁路路基碳排放测算及关键降碳因素分析[J].铁道学报，2025,47(9):212-222.Shen Zhongshuai,Bao Xueying,Liu Bei sheng,et al.Measurement of carbon emissions from railway subgrade and analysis of key carbon reduction factors[J].Journal of the China Railway Society,2025,47(9):212-222.
• [23] IEEE Std 1474.1-2004 Communications-based train control (cbtc) performance and func-tional requirements[S].
• [24] Sharma S R,Singh B,Kaur M.A novel approach of ensemble methods using the stacked generalization for high-dimensional datasets[J].IETE journal of research,2023,69(10):6802-6817.
• [25] Federal Transit Administration (FTA).National transit database (NTD) safety & security reporting policy manual[EB/OL].(2026-05-19)[2026-09-29].https://www.transit.dot.gov/ntd/2026-ntd-safety-and-security-reporting-policy-manual-0.
• [26] 国务院办公厅.国家城市轨道交通运营突发事件应急预案[EB/OL].(2015-05-14)[2026-09-29].https://www.gov.cn/zhengce/content/2015-05/14/content_9751.htm.State Council General Office.Issuing the national emergency plan for urban rail transit operational emergencies[EB/OL].(2015-05-14)[2026-09-29].https://www.gov.cn/zhengce/content/2015-05/14/content_9751.htm.
• [27] Yan J,Liu Y,Meng H,et al.Wind turbine generator early fault diagnosis using LSTM-based stacked denoising autoencoder network and stacking algorithm[J].International Journal of Green Energy,2024,21(11):2477-2492.
• [28] Hou H,Chen X,Li M,et al.Prediction of user outage under typhoon disaster based on multi-algorithm stacking integration[J].International Journal of Electrical Power & Energy Systems,2021,131:107123.
• [29] Schl??gl M,Stütz R,Laaha G,et al.A comparison of statistical learning methods for deriving determining factors of accident occurrence from an imbalanced high resolution dataset[J].Accident Analysis & Prevention,2019,127:134-149.
• [30] Apley D,Zhu J.Visualizing the effects of predictor variables in black box supervised learning models[J].Journal of the Royal Statistical Society Series B:Statistical Methodology,2020,82(4):1059-1086.