基于事故综合强度的公交事故严重程度分析

为更加精确地对公交事故严重程度进行分类以探究其影响因素，本文提出一种基于事故综合强度+K-means的公交事故严重程度分类方法，并基于此分类方法建立公交事故严重程度影响因素分析模型。首先，针对传统事故严重程度分类中的定性分类方法，引入事故综合强度法定量计算公交事故严重程度，并运用K-means聚类算法对事故严重程度进行聚类。其次，选取环境、驾驶员、道路车辆和事故特征这 4 方面的 17 个因素作为自变量，分别将事故综合强度+K-means分类法和传统分类法的结果作为因变量，运用有序Logit模型分析公交车事故严重程度，同时利用平均边际效应量化各显著因素的影响程度，以佛山市2021年156起公交车事故数据为例进行分析。结果表明，基于事故综合强度+K-means分类法的有序Logit模型具有更好的拟合优度。高峰期、换道、超速、加速度过大、注意力分散和进出站会增大发生极严重公交车事故的概率，增大的概率分别为11.57%、29.06%、23.98%、17.13%、30.97%和12.27%；白天和晴天会减小发生极严重公交车事故的概率，减少的概率分别为22.31%和12.34%。

Bus Accident Severity Analysis Based on Comprehensive Accident Intensity

To classify the severity of bus accidents more accurately and identify the influence factors for the bus accidents, this paper proposed a classification method based on comprehensive accident intensity and K-means algorithm. An analysis model of the influence factors of bus accident severity was also developed based on the results of the classification method. Comparing to the traditional qualitative classifications of accident severity, the comprehensive accident intensity method was introduced to calculate the bus accident severity, and the accident severity was classified by the K-means clustering algorithm. Then, 17 factors from environment, drivers, roads/vehicles and accident characteristics were selected as independent variables. The results of comprehensive accident intensity & K- means classification method and traditional four classification method were used as the dependent variables. The ordered Logit model was used to analyze the bus accident severity. In addition, the average marginal effect was used to quantify the impact of each significant factor and the bus accident data of Foshan City in 2021 was analyzed as an example. The results show that the ordered Logit model based on the classification method of comprehensive accident intensity and K-means algorithm has superior statistical performance. Peak periods, lane changes, speeding, excessive acceleration distracted driving, and pulling in and out of stations will increase the probability of serious bus accidents by 11.57%, 29.06%, 23.98%, 17.13%, 30.97% and 12.27%, respectively. Daytime and sunny days respectively reduce the probability of a serious bus accident by 22.31% and 12.34%.