贝叶斯网络在地铁运营事故严重程度预测中的应用研究  

Research on subway operation accident severity warning based on Bayesian network

作　　者：雷斌 田伯轩 李佳晨 马谦 刘鑫 LEI Bin;TIAN Boxuan;LI Jiachen;MA Qian;LIU Xin(College of Civil Engineering,Xi'an University of Architecture and Technology,Xi'an 710055,China;Northwest Electric Power Design Institute Co.,Ltd.of China Power Engineering Consulting Group,Xi'an 710075,China;Xi'an Zhongtie Railway Transportations Co.,Ltd.,Xi'an 710038,China)

机构地区：[1]西安建筑科技大学土木工程学院,西安710055 [2]中国电力工程顾问集团西北电力设计院有限公司,西安710075 [3]西安中铁轨道交通有限公司,西安710038

出　　处：《安全与环境学报》2024年第7期2751-2758,共8页Journal of Safety and Environment

基　　金：陕西省科学技术厅社会发展领域项目(2021SF-486)。

摘　　要：为了针对不同地铁运营事故致因预测事故后果的严重程度,将贝叶斯网络应用于地铁运营事故严重程度预警,建立地铁运营事故严重程度预测模型。首先,利用树型朴素贝叶斯算法对地铁运营事故数据进行模拟训练,得到不同事故致因的重要度;其次,通过贝叶斯网络分析和网络推理得到各事故致因的概率分布;最后,以搜集得到的国内2001—2022年地铁运营事故数据为例开展实例分析,并选取准确率、召回率和精确率等指标验证模型预测结果的有效性。结果表明:机器学习获得的网络结构显示车辆是否故障、工作人员是否操作失误、人员因素其他原因是影响事故严重程度的重要因素;通过对模型的预测结果进行分析统计,得到不同事故致因对地铁事故严重程度的影响,可以为地铁管理部门进行事故预警提供智能辅助决策,为事故救援处置提供行动依据。To predict the severity of accident consequences for different causes of subway operation accidents,a Bayesian network is applied to the early warning of the severity of subway operation accidents and establish a prediction model.The model aims to determine the severity of subway operation accidents.Firstly,by referring to the relevant literature on rail traffic accidents and the authoritative data of relevant departments,the data of domestic subway operation accidents from 2001 to 2022 were collected and randomly divided into 70%training set and 30%test set.After imbalanced processing of the training set data,simulation training was carried out to obtain the importance of different accident-causing attributes.Secondly,by Bayesian network analysis and network reasoning obtain the probability distribution of each causal attribute node.Finally,the model was evaluated using accuracy rate,precision rate,and recall rate.It is found that the imbalanced processing of the training set data leads to an improvement in the results of each test index,demonstrating the necessity of addressing data imbalance.The indexes of the test set can reach more than 85%.Through the statistics of the recall rate and precision rate of the prediction results of major accidents,it is found that the comprehensive evaluation index F 3 of the test set can also reach 91.5%and 90.1%,verifying the prediction performance of the model.The results indicate that the structure of the network obtained through machine learning demonstrates that the severity of accidents is influenced by various factors,including vehicle faults and personnel-related factors such as incorrect operation.Through the analysis and statistics of the prediction results of the model,the impact of different accident causes on the severity of subway accidents is obtained.This information can assist the subway management department in making intelligent decisions for early accident warning at the beginning of the accident and provide a basis for accident rescue and disposal.

关 键 词：安全社会科学 地铁事故预测 贝叶斯网络 事故严重程度 

分 类 号：X928.03[环境科学与工程—安全科学]