重大道路交通事故严重程度影响因素分析  

Influencing Factors on Severity of Serious Road Traffic Accidents

作　　者：白玉[1] 魏毅 BAI Yu;WEI Yi(Key Laboratory of Road and Traffic Engineering of the Ministry of Education,Tongji University,Shanghai 201804,China)

机构地区：[1]同济大学道路与交通工程教育部重点实验室,上海201804

出　　处：《交通与运输》2022年第3期22-26,共5页Traffic & Transportation

基　　金：国家自然科学基金项目(71871163);中国交通建设股份有限公司科技研发项目(2019-ZJKJ-ZDZX02)。

摘　　要：基于事故数据的道路交通事故严重程度影响因素分析是制定交通安全对策的基础,不同类型、算法的建模方法对事故严重程度影响因素的拟合精度不同,提取的显著影响因素存在差异。为提高研究结果准确性,以某地区人员伤亡和重大财产损失的道路交通事故信息为数据集,运用Multinomial Logit、决策树、神经网络3类方法分别建模分析事故严重程度影响因素。鉴于3类模型输出的显著变量存在差异,通过交叉对比验证进行显著影响因素判别。研究发现,事故认定原因、事故发生纵向位置、事故发生道路类型等是事故严重显著影响因素,研究成果有效修正了单一模型的偏差,对精准制定事故改善对策具有指导意义。The analysis of influencing factors of the severity of road traffic accidents based on accident data is the basis for formulating traffic safety measures.Different types of modeling methods and algorithms have different fitting accuracy for influencing factors of accident severity,and extract different significant influencing factors.To improve the accuracy of the results,multinomial logit,decision tree and neural network are used to model and analyze the factors affecting the severity of road traffic accidents.In view of the differences in significant variables among the outputs of the three types of models,cross contrast verification is used to identify significant influencing factors.The results show that the cause,the longitudinal location of the accident and the type of road are the significant influencing factors of the accident severity.The results effectively correct the deviation of the single model,and have guiding significance for the accurate formulation of the accident improvement measures.

关 键 词：道路交通事故 严重程度影响因素 Multinomial Logit模型 决策树 神经网络 

分 类 号：U491[交通运输工程—交通运输规划与管理]