摘要
为深入剖析交通事故严重程度影响因素,实现交通事故主动预防、精准防控,选择车、路、环境等属性中显著性较高因素作为输入变量、以事故严重程度作为二分类输出变量,构建交通事故严重程度影响因素CART决策树模型,并选择CCP算法对模型进行优化求解。选择相对误差值、ROC曲线和AUC等作为评价指标验证模型的有效性,模型相对误差值仅为6.08%,AUC值为0.93。研究结果表明:在道路及交通环境方面,恶劣天气、塌陷路面、道路存在隐患、交叉口无信号控制等因素对交通事故严重程度影响较大;在车辆方面,大型货车发生较大及以上事故概率是其他车辆的4倍,上述因素应作为交通事故防控的重点。
In order to deeply analyze the influencing factors of traffic accident severity and realize the active prevention and accurate prevention and control of traffic accidents,the factors with high significance in the attributes such as vehicle,road and environment were selected as the input variables and the accident severity was taken as the secondary classification output variable.The CART decision tree model of the influencing factors of traffic accident severity was established and the CCP algorithm was selected to optimize and solve the proposed model.The relative error value,ROC curve and AUC were selected as evaluation indexes to verify the effectiveness of the proposed model.The relative error value of the proposed model was only 6.08%and the AUC value was 0.91.The research results show that in terms of road and traffic environment,the factors such as bad weather,collapsed pavement,hidden dangers in the road and no signal control at the intersection have a great impact on the severity of traffic accidents;in terms of vehicles,the probability of major accidents or above of large trucks is 4 times that of other vehicles.Therefore,the above factors should be the focus of traffic accident prevention and control.
作者
张萌萌
李泽文
徐云帆
刘永栋
ZHANG Mengmeng;LI Zewen;XU Yunfan;LIU Yongdong(School of Transportation and Logistics Engineering,Shandong Jiaotong University,Ji nan 250357,Shandong,China)
出处
《重庆交通大学学报（自然科学版）》
CAS
CSCD
北大核心
2023年第9期93-98,共6页
Journal of Chongqing Jiaotong University(Natural Science)
基金
国家自然科学基金项目(52102412)
全国统计科学研究项目(2021LY017)
2021年度山东省自然科学基金项目(ZR202103040503)
2021年度山东省自然科学基金项目(ZR2021QF110)。
关键词
交通运输工程
交通安全
优化CART
交通事故严重程度
CCP算法
traffic and transportation engineering
traffic safety
optimized CART
traffic accident severity
CCP algorithm
