Research on subway operation accident severity warning based on Bayesian network
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 F3 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.
safety social sciencesubway accident predictionBayesian networkseverity of accident
NETL
NSTL
万方数据
