基于贝叶斯网络的高速公路交通事故严重程度预测及致因分析

Prediction and Cause Analysis of Freeway Traffic Accident Severity Based on Bayesian Network

为了深入研究高速公路交通安全,剖析高速公路交通事故的发生机理以及各类因素对高速公路交通事故严重程度的影响,收集曲靖市境内沪昆高速段2017—2019年的1 939起交通事故进行研究。以事故严重程度为因变量,筛选出人、车、路、环境4个大类下的与事故严重程度相关的19个影响因素为自变量,采用数据融合法基于树增广型贝叶斯网络构建事故严重程度预测模型,量化各因素间的影响关系,经特征筛选找出关键致因,并结合案例进行推理分析。结果表明:影响高速公路交通事故严重程度的关键致因依次为天气情况、视距情况、路面情况等。模型对高速公路事故严重程度预测准确率可达84.27%,高于传统贝叶斯方法,模型有效性验证良好。针对事故主要致因提出改进建议,可为交管部门提供准确事故信息辅助决策,加快事故响应速度,提高事故应急指挥能力。

In order to study the freeway traffic safety in depth,the mechanism of freeway traffic accidents and the influence of various factors on the severity of freeway traffic accidents were analyzed.1939 traffic accidents on Shanghai-Kunming Expressway in Qujing City from 2017 to 2019 were collected for research.Taking the accident severity as dependent variable,19 influencing factors related to the accident severity were selected as independent variables under four categories such as people,vehicle,road and environment.Based on tree-enhanced Bayesian network,a prediction model of accident severity was established by data fusion method,and the influence relationships among various factors were quantified.The key causes were found by feature screening,and the reasoning analysis based on case studies was carried out.The results show that the key factors affecting the severity of traffic accidents on freeways are successively weather conditions,sight distance conditions,road conditions,etc.The prediction accuracy of the proposed model for the severity of freeway accidents can reach 84.27%,which is higher than that of the traditional Bayesian method,and the validity of the proposed model is well verified.According to the main causes of the accident,some improvement suggestions are put forward,which can provide accurate accident information to assist decision-making for traffic management departments,speed up the accident response and improve the emergency command ability.