基于改进Apriori的地铁运维危险源致灾度量化及风险预判  

Disaster-Causing Degree Quantification and Risk Prediction of Hazard Sources in Metro Operation and Maintenance Based on Improved Apriori

作　　者：唐永升 李花子 TANG Yongsheng;LI Huazi(School of Management,Shanghai University of Engineering Science,Shanghai 201620,China;Academic Affairs Office,Shanghai University of Engineering Science,Shanghai 201620,China;International Institute of Creative Design,Shanghai University of Engineering Science,Shanghai 201620,China)

机构地区：[1]上海工程技术大学管理学院,上海201620 [2]上海工程技术大学教务处,上海201620 [3]上海工程技术大学国际创意设计学院,上海201620

出　　处：《铁道运输与经济》2025年第4期152-161,共10页Railway Transport and Economy

基　　金：上海市哲学社会科学规划课题(2024BGL005)。

摘　　要：为精准预判地铁运营事故严重程度,提出有序约束Apriori-RF算法量化运营事故灾害度等级。首先,以人员伤亡、列车延误和设施损坏3个维度构建致灾量化模型,运用K-means算法聚类成4个灾害度等级;其次,引入有序约束改进Apriori算法,挖掘风险与灾害度等级间的非线性关系,得到42条有效关联规则;再次,将其输入随机森林算法进行训练,通过基尼系数得到灾害度等级的风险重要度;最后,采用有序约束Apriori-RF方法与随机森林算法作实例验算并对比。研究表明:Apriori-RF可使关联规则挖掘有效度提升74.9%,且效率更高;结果的均方根误差(RMSE)降低14%、加权均方根误差(WRMSE)降低36%,表明其准确度也得到显著提升。研究成果可为量化预判地铁运营事故的灾害度等级提供一种精确且有效的方法,对保障地铁运营安全及事故减灾防控有理论意义和应用价值。To accurately predict the severity of metro operation accidents,this study proposed an ordered constraint Apriori-RF algorithm to quantify disaster degree levels of operation accidents.Firstly,a disaster-causing quantification model was constructed based on three dimensions:casualties,train delays,and facility damages.The K-means algorithm was clustered into four disaster degree levels.Secondly,the improved ordered constraint Apriori algorithm was introduced to explore nonlinear relationships between risks and disaster degree levels,yielding 42 effective association rules.Thirdly,these rules were input into a random forest algorithm for training,obtaining risk importance for disaster degree levels through Gini coefficient analysis.Finally,cases were verified and compared through the ordered constraint Apriori-RF method and random forest algorithm.Research demonstrates that the Apriori-RF method improves association rule mining effectiveness by 74.9%with higher efficiency.The results show a 14%reduction in the Root Mean Square Error(RMSE)and a 36%decrease in the Weighted Root Mean Square Error(WRMSE),indicating significantly higher accuracy.The research findings provide an accurate and effective method for quantitatively predicting disaster degree levels in metro operation accidents,holding theoretical significance and practical value in ensuring operation safety and disaster mitigation and prevention.

关 键 词：地铁运维安全 风险管控 有序约束Apriori-RF方法 致灾度量化 风险预判 

分 类 号：U293.6[交通运输工程—交通运输规划与管理]