An Analysis of The Impact Factors of Head Injuries of Two-wheeler Riders Using a Latent Class Logit Model
-
摘要: 为研究汽车-两轮车碰撞事故中骑行者头部伤害的影响机理,探究头部损伤致因关系中存在的模式差异与异质性,基于中国交通事故深入研究(CIDAS)数据库中的2 806起两轮车事故,建立潜类别Logit模型。以骑行者头部伤害严重程度为因变量,将驾驶人、车辆、道路、环境和事故碰撞特征的相关因素作为自变量,取显著水平为0.05,建立多项式Logit模型,在此基础上根据拟合优度指标确定最优分组数并构建出潜类别Logit模型。研究结果表明:模型将事故样本划分为2个类别群体,2个群体在参数、变量分布特征和预测概率上存在显著差异,当事故样本具有“两轮车初始速度>30 km/h”、“骑行者碰撞后抛出距离>10 m”等特征时易被归类到类别1,且类别1对应于更严重的头部伤害;骑行者年龄>50岁、汽车类型为商用货车、两轮车类型为摩托车、事故发生在市区外、两轮车初始速度>30 km/h、头部碰撞玻璃和抛出距离>5 m都会增加头部损伤的严重程度;汽车驾驶人行驶意图为停车或变道时,存在造成严重两轮车事故的风险;佩戴头盔会减弱骑行者头部受到的伤害。
-
关键词:
- 交通安全 /
- 汽车-两轮车碰撞事故 /
- 骑行者 /
- 头部伤害 /
- 潜类别Logit模型 /
- 异质性
Abstract: This paper studies the impact factors of head injuries of two-wheeler riders via a novel latent class Logit model. The seriousness of the head injuries of the riders is used as the dependent variable, while the factors of drivers, vehicles, roads, environment, and characteristics of collisions are taken as independent variables. A multinomial Logit model is developed with a significance level of 0.05. On this basis, the optimal number of classes is determined according to the goodness of fit. A latent class Logit model is developed based on 2806 two-wheeler collision data collected by the China In-depth Accident Study (CIDAS). According to the results, the model divides accident samples into two distinct categories. The two groups differ significantly in terms of parameter values, variable distribution characteristics, and the likelihood of predicting the outcome. Specifically, accidents with characteristics such as"two-wheelers initial speed is greater than 30 km/h"and"throwing distance is greater than 10 meters"are more likely to be classified as Class 1, which refers to the riders with more severe head injuries. In addition, severer head injuries are likely to occur under the following scenarios: including when a rider is over fifty, the colliding vehicle is a commercial truck, the two-wheeler is a motorcycle, the accident occurs outside a city, the two-wheeler is traveling above 30 km/h, the head collides with the glass, and the distance to the collision site after the collision is greater than 5 meters. Moreover, the risk of serious two-wheeler collisions is higher when a car driver intends to park or change lanes. Helmets are shown to reduce head injuries among riders.-
Key words:
- traffic safety /
- car to two-wheeler collision /
- rider /
- head injury /
- latent class Logit model /
- heterogeneity
-
表 1 自变量描述性统计
Table 1. Description statistics of independent variables
变量 变量描述 频数(占比/%) 变量 变量描述 频数(占比/%) 事故碰撞特征 汽车驾驶人年龄段/岁 >18~30* 951(33.89) 汽车驾驶人是否采取避让措施 否* 1 760(62.72) >30~40 924(32.93) 是 1 046(37.28) >40~50 659(23.49) 汽车驾驶人碰撞前行驶意图 保持直行* 1 465(52.21) >50~60 228(8.13) 右转 501(17.85) >60 44(1.57) 左转 468(16.68) 骑行者年龄段/岁 ≤30* 517(18.42) 其他(驻车、变道等) 372(13.26) >30~40 400(14.26) 汽车碰撞前是否减速 否* 1 862(66.36) >40~50 640(22.81) 是 944(33.64) >50~60 742(26.44) 两轮车碰撞前初始速度(/km/h) ≤20* 1 114(39.70) >60 507(18.07) >20~30 922(32.86) 汽车驾驶人驾龄/年 ≤5* 1 137(40.52) >30 770(27.44) >5~10 731(26.05) 汽车行驶速度(/km/h) ≤30* 1 468(52.32) >10 917(32.68) >30~60 1 073(38.24) 无证驾驶 21(0.75) >60 265(9.44) 车辆特征 骑行者碰撞后抛出距离S/m ≤2* 631(22.49) 汽车内部是否有视线障碍 否* 2 665(94.98) >2~5 859(30.61) 是 141(5.02) >5~10 657(23.41) 汽车车辆类型 乘用车及小型客车* 2 323(82.79) >10 659(23.49) 商用货车 483(17.21) 骑行者碰撞后抛射特征 无抛射* 182(6.49) 两轮车车辆类型 电动车* 1 392(49.61) 滑滚 1 044(37.21) 摩托车 1 178(41.98) 翻滚 960(34.21) 自行车 236(8.41) 小碰撞面 149(5.31) 道路特征 大碰撞面 167(5.95) 路面材料 沥青* 1 959(69.81) 其他(碾压等) 304(10.83) 水泥 821(29.26) 骑行者头部碰撞类型 无* 2 156(76.84) 其他(沙砾等) 26(0.93) 未变形或轻微划痕 110(3.92) 路灯情况 路灯关闭* 1 754(62.51) 有变形 151(5.38) 路灯开启 555(19.78) 碰撞玻璃造成损坏 337(12.01) 无路灯 497(17.71) 其他(碾压等) 52(1.85) 环境特征 驾驶人特征 汽车外部视线障碍 无* 2 304(82.11) 汽车驾驶人性别 女* 373(13.29) 短暂的视线障碍 294(10.48) 男 2 433(86.71) 持续的视线障碍 208(7.41) 骑行者性别 女* 756(26.94) 事故地点所属区域 市区外* 1 395(49.71) 男 2 050(73.06) 市区内 1 411(50.29) 骑行者是否佩戴头盔 否* 2 309(82.29) 天气条件 晴天* 2 118(75.48) 是 497(17.71) 阴天或多云 600(21.38) 雾天等 88(3.14) 注:*为参照组。 表 2 “汽车驾驶人驾龄”虚拟变量的定义
Table 2. Definition of the dummy variable " car driver's driving age"
汽车驾驶人驾龄X /年 虚拟变量 X1 X2 X3 ≤5* 0 0 0 >5~10 1 0 0 >10 0 1 0 无证驾驶 0 0 1 表 3 模型方法比较
Table 3. Comparison of modeling methods
模型 分类 原理 优点 局限 决策树、随机森林、XGBoost、LightGBM等 机器学习模型 通过对数据的1个子集(训练集)进行训练,得到1个模型,并利用另1个子集(测试集)来判断所建立模型的准确性 预测精度高,易于处理大量复杂数据,常用于事故预测 可解释性较差;使用采样法来调整数据平衡时会改变原始数据的分布,研究样本只是理论上存在 潜类别Logit模型 离散选择模型 建模目的是找到变量之间的关系,并确定关系的显著性。基于随机效应和效用最大化理论,通过显著性检验等方法来检验模型的准确性 可解释性好,能够识别群体层面的异质性,常用于事故严重程度影响因素分析 分类数增多时,会使模型参数过多,导致模型不收敛 表 4 潜类别Logit模型参数估计结果
Table 4. Parameter estimation results of latent class Logit model
变量 多项式Logit模型 潜类别Logit模型 类别1 类别2 参数估计值 P值 参数估计值 P值 参数估计值 P值 重伤或致死 骑行者年龄>50~60岁 0.348 0.007 2.669 0.012 骑行者年龄>60岁 0.383 0.010 0.681 0.010 汽车驾驶人无证驾驶 1.069 0.029 骑行者佩戴头盔 -0.542 0.001 -1.026 0.000 汽车为商用货车 1.017 0.000 3.090 0.006 0.792 0.000 两轮车为摩托车 0.362 0.005 0.618 0.001 事故地点在市区内 -0.374 0.001 -0.715 0.000 汽车驾驶人碰撞前行驶意图为右转 -0.655 0.001 -0.778 0.003 汽车驾驶人碰撞前行驶意图为驻车、变道等 -0.569 0.001 3.568 0.009 -1.611 0.000 汽车碰撞前减速 -0.375 0.002 -0.707 0.000 两轮车碰撞前初始速度V0 >30 km/h 0.457 0.001 2.198 0.070 骑行者碰撞后抛出距离>5~10 m 0.548 0.000 3.397 0.012 骑行者碰撞后抛出距离>10 m 0.823 0.000 4.064 0.011 0.513 0.051 骑行者头部碰撞玻璃 2.160 0.000 9.095 0.002 1.651 0.000 轻度或中度伤害 常数项 1.785 0.000 5.970 0.013 1.467 0.000 骑行者为男性 -0.227 0.018 -1.549 0.053 骑行者佩戴头盔 -0.442 0.000 -2.929 0.015 汽车存在车内视线障碍 -0.603 0.005 -0.668 0.072 无路灯 0.375 0.000 2.378 0.058 两轮车碰撞前初始速度V0 >30 km/h 0.377 0.000 6.808 0.001 0.796 0.010 汽车行驶速度>30~60 km/h -0.374 0.008 4.198 0.012 -1.341 0.000 汽车行驶速度>60 km/h -1.138 0.000 -1.997 0.000 骑行者头部碰撞未变形或轻微划痕 -0.620 0.011 骑行者头部碰撞有变形 -1.369 0.000 -10.121 0.010 骑行者头部碰撞玻璃 1.445 0.000 1.845 0.000 骑行者头部受到碾压等 -0.927 0.019 无伤害 常数项 2.497 0.000 9.906 0.001 1.710 0.000 骑行者年龄>60岁 -0.480 0.000 -2.298 0.075 汽车行驶速度>30~60 km/h -0.681 0.000 -0.601 0.003 汽车行驶速度>60 km/h -1.643 0.000 -1.695 0.000 骑行者碰撞后抛出距离>10 m -0.441 0.001 -2.310 0.085 骑行者头部碰撞未变形或轻微划痕 -2.438 0.000 -5.688 0.027 -2.382 0.000 骑行者头部碰撞有变形 -3.518 0.000 -6.353 0.008 -3.904 0.000 骑行者头部受到碾压等 -1.246 0.000 -7.698 0.006 模型拟合优度指标 Likelihood ratio(LR)chi-square 990.588 1 319.443 Prob > chi-square 0.000 0.000 AIC(Akaike information criterion) 5 009.5 4 984.0 McFadden R2 0.169 0.214 注:1表中只列出了潜类别Logit模型中P值< 0.1的参数估计值;
2AIC值越小表示模型拟合越好;McFadden R2值在0~1之间,越大表示模型拟合越好。表 5 相关结果与应用建议对应表
Table 5. Comparison table of relevant results and application suggestions
相关结果* 应用建议 >50岁的骑行者头部受更高等级伤害概率增加 开展50岁以上骑行者佩戴头盔的专项治理与引导** 男性骑行者头部受伤害的概率更低 开展针对女性群体佩戴头盔的专项宣传** 汽车驾驶人行驶意图为驻车、变道等时在不同类别中对应的头部伤害存在差异 提高汽车驾驶辅助安全系统(包括泊车系统)安装比例 不同类别群体中头盔的保护效果存在差异 根据头围、头部轮廓等人体特征来研制并提供头盔;将自行车骑行者的头盔纳入监管范围** 商用货车造成头部受重伤或致死伤害的概率增加 提高商用货车盲区监测系统配备率 摩托车参与事故中骑行者头部受重伤或致死伤害的概率更大 加强对摩托车骑行者佩戴头盔的监管 两轮车撞前初始速度>30 km/h会导致更加严重的头部伤害 进一步研究考虑两轮车骑行者伤害的合理限速区间 汽车低速情况下对两轮车骑行者造成的伤害未必就小 开发涵盖低速运行危险场景特征的汽车辅助安全系统 市区内发生重伤或致死两轮车事故的概率低 郊区与城乡结合部应是预防工作的重点区域,提高上述区域交通设施、照明设施等完备率 骑行者撞后抛出距离的增加会带来更严重的头部伤害 加强汽车前部缓冲保护和两轮车前端缓冲保护设计,减少碰撞能量和抛出距离** 骑行者头部碰撞玻璃会引发极为严重的事故后果 汽车前部结构和缓冲保护设计中尽量避免骑行者头部与挡风玻璃发生碰撞** 注:*相关结果为全部或部分类别中发现的特性;**为论文通过头部损伤研究提出的新建议,其他条目则是与以往文献相符的建议。 -
[1] CHEN Q, DAI B. Characteristics analysis of two-wheeler accidents in China based on CIDAS data between 5 year[C]. The 14th International Forum of Automotive Traffic Safety, Changsha: State Key Laboratory of Advanced Design and Manufacturing for Vehicle Body, Hunan University, 2017. (in Chinese) [2] 国家统计局. 中国统计年鉴[M]. 北京: 中国统计出版社, 2021.National Bureau of Statistics. China Statistical Yearbook[M]. Beijing: China Statistics Press, 2021. (in Chinese) [3] 王兴华, 彭勇. 汽车-自行车/摩托车碰撞事故中骑车人头腿部动力学响应对比研究[J]. 振动与冲击, 2018, 37(1): 156-162. doi: 10.13465/j.cnki.jvs.2018.01.024WANG X H, PENG Y. Comparative investigation on the head and leg dynamic responses of two-wheeler riders in vehicle-bicycle/motorcycle accidents[J]. Journal of Vibration and Shock, 2018, 37(1): 156-162. (in Chinese) doi: 10.13465/j.cnki.jvs.2018.01.024 [4] GAO W, BAI Z, ZHU F, et al. A study on the cyclist head kinematic responses in electric-bicycle-to-car accidents using decision-tree model[J]. Accident Analysis & Prevention, 2021 (160): 106305. [5] 胡林, 方胜勇, 黄晶, 等. 基于逻辑回归的二轮车-汽车碰撞事故深度分析[J]. 汽车工程, 2016, 38(11): 1288-1293. doi: 10.3969/j.issn.1000-680X.2016.11.002HU L, FANG S Y, HUANG J, et al. In-depth analysis on cycle-vehicle crash accident based on logistic regression[J]. Automotive Engineering, 2016, 38(11): 1288-1293. (in Chinese) doi: 10.3969/j.issn.1000-680X.2016.11.002 [6] BLACKMAN R A, HAWORTH N L. Comparison of moped, scooter and motorcycle crash risk and crash severity[J]. Accident Analysis & Prevention, 2013(57): 1-9. [7] ROBARTES E, CHEN T D. The effect of crash characteristics on cyclist injuries: An analysis of Virginia automobile-bicycle crash data[J]. Accident Analysis & Prevention, 2017(104): 165-173. [8] PENG Y, CHEN Y, YANG J, et al. A study of pedestrian and bicyclist exposure to head injury in passenger car collisions based on accident data and simulations[J]. Safety Science, 2012, 50(9): 1749-1759. doi: 10.1016/j.ssci.2012.03.005 [9] MANNERING F L, SHANKAR V, BHAT C R. Unobserved heterogeneity and the statistical analysis of highway accident data[J]. Analytic Methods in Accident Research, 2016(11): 1-16. [10] AFGHARI A P, HEZAVEH A M, HAQUE M M, et al. A home-based approach to understanding seatbelt use in single-occupant vehicles in Tennessee: application of a latent class binary logit model[J]. Accident Analysis & Prevention, 2020(146): 105743. [11] CERWICK D M, GKRITZA K, SHAHEED M S, et al. A comparison of the mixed logit and latent class methods for crash severity analysis[J]. Analytic Methods in Accident Research, 2014(3): 11-27. [12] ADANU E K, HAINEN A, JONES S. Latent class analysis of factors that influence weekday and weekend single-vehicle crash severities[J]. Accident Analysis & Prevention, 2018 (113): 187-192. [13] 施颖, 潘义勇, 吴静婷. 基于随机参数Logit模型的校车事故伤害严重程度分析[J]. 交通信息与安全, 2021, 39(5): 43-49. doi: 10.3963/j.jssn.1674-4861.2021.05.006SHI Y, PAN Y Y, WU J T. An analysis of injury severities in school bus accidents based on random parameter logit models[J]. Journal of Transport Information and Safety, 2021, 39 (5): 43-49. (in Chinese) doi: 10.3963/j.jssn.1674-4861.2021.05.006 [14] XIE Y, ZHAO K, HUYNH N. Analysis of driver injury severity in rural single-vehicle crashes[J]. Accident Analysis & Prevention, 2012(47): 36-44. [15] GREENE W H, HENSHER D A. A latent class model for discrete choice analysis: Contrasts with mixed logit[J]. Transportation Research Part B: Methodological, 2003, 37(8): 681-698. [16] YU H, LI Z, ZHANG G, et al. A latent class approach for driver injury severity analysis in highway single vehicle crash considering unobserved heterogeneity and temporal influence[J]. Analytic Methods in Accident Research, 2019 (24): 100110. [17] LI Z, WU Q, CI Y, et al. Using latent class analysis and mixed logit model to explore risk factors on driver injury severity in single-vehicle crashes[J]. Accident Analysis & Prevention, 2019(129): 230-240. [18] 王卫杰, 沈轩霆, 王贵彬, 等. 电动自行车骑行者事故伤害程度影响因素分析[J]. 中国安全科学学报, 2019(2): 20-25.WANG W J, SHEN X T, WANG G B, et al. Analysis of factors affecting injury to electric bicycle rider in crash[J]. China Safety Science Journal, 2019(2): 20-25. (in Chinese) [19] 朱彤, 朱秭硕, 张杰, 等. 高峰期电动二轮车驾驶人事故伤害致因及随机参数分析[J]. 安全与环境学报, 2022, 22 (1): 271-280. https://www.cnki.com.cn/Article/CJFDTOTAL-AQHJ202201039.htmZHU T, ZHU Z S, ZHANG J, et al. Analysis of electric bicyclist injury severity during traffic peak period with random parameters[J]. Journal of Safety and Environment, 2022, 22 (1): 271-280. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-AQHJ202201039.htm