摘要
为提高电动自行车交通安全管理水平,基于Logistic回归与树型贝叶斯网络(Tree Augmented Naive Bayes,TAN)的组合方法,探究电动自行车交通事故严重程度的影响因素。首先,收集某市2016~2020年电动自行车交通事故数据,将事故严重程度作为因变量,电动自行车驾驶人年龄等类别属性作为自变量,构建二分类Logistic回归模型;其次,根据回归模型结果,选择显著的自变量和影响因素,在TAN模型中进行单一证据变量以及多证据变量耦合推理分析,量化其影响大小,并分析数据中的异质性。结果表明,11个因素对因变量有显著影响,其中“大中型车辆”是最重要的影响因素;“右转”和“追尾碰撞”两个因素对因变量具有异质影响;电动自行车与右转的大中型车辆发生追尾碰撞的死亡事故概率最高,达到81.1%。
To improve the level of traffic safety management for electric bicycle,the influencing factors of the traffic accidents severity for electric bicycles were investigated based on the combination method of Logistic regression and tree augmented naive bayes.Firstly,the traffic accidents data of electric bicycle in a city from 2016 to 2020 was collected,and a binary Logistic regression model was constructed with the accident severity as the dependent variable and the age of the electric bicycle rider and other categorical variables as independent variables.Secondly,based on the results of the regression model,significant independent variables and influencing factors were selected,and the reasoning analysis of single evidence variable and multiple evidence variables was carried out in TAN model to quantify their influence and analyze the heterogeneity in the data.The results show that 11 factors have significant influence on the dependent variable,among which“large and medium-sized vehicle”is the most important factor.Two factors,“right turn”and“rear-end collision”,have heterogeneous effects on the dependent variable.The probability of death in rear-end collisions between e-bikes and right-turning large and medium-sized vehicles is the highest,reaching 81.1 percent.
作者
柯星安
丁立民
赵丹
KE Xingan;DING Limin;ZHAO Dan(School of Traffic Management,People s Public Security University of China,Beijing 100038,China)
出处
《中国人民公安大学学报（自然科学版）》
2023年第2期47-54,共8页
Journal of People’s Public Security University of China(Science and Technology)
基金
2022年中国人民公安大学公共安全行为科学与工程科技创新项目(2022KXGCKJ06)。
