An Analysis of Occupant Death Risk of 5-Seater Cars in Two-vehicle Collisions
-
摘要: 为探究5座乘用车乘员在2车碰撞事故下的死亡风险，研究了6种单一特征变量对乘员致死率的影响，进而基于二项Logistic回归模型分别对单一特征变量和组合特征变量进行显著性分析。通过9种常用的分类算法，结合网格搜索的调参方法，以F1为衡量指标选出相对较优的3种分类算法，即投票分类器、梯度提升及决策树，来构建多特征组合下的死亡风险预测模型。研究结果表明：①单一特征变量中行驶方向、路段类型、碰撞对象、乘坐位置对乘员死亡有显著影响。其中，异向行驶的车辆碰撞与同向行驶相比，乘员的死亡风险增加72%；非高速交叉路段与高速路段相比，乘员的死亡风险降低69%；碰撞对象为商用货车、商用客车的乘员死亡风险分别是乘用车的5倍和3倍，若在非高速非交叉路段发生碰撞则乘员死亡风险升至8倍左右，若在高速路段则高达15倍左右；相对于驾驶位乘员，副驾驶位乘员的死亡风险增加70%，且该位置乘员的死亡风险在高速路段会升高到驾驶位乘员的近4倍；②碰撞对象和路段类型是影响乘员死亡情况的主要特征变量；③由模型的预测结果可知：5座乘用车的正面或后面与商用货车在高速路段或非高速非交叉路段发生碰撞，乘员的死亡风险高于生存几率。Abstract: To study occupant death risk of 5-seater cars in two-vehicle collisions, the impacts of six variables on fatality rates are compared. Moreover, binary logistic regression is used to analyze the influences of different features and their combinations on occupant death risk. A model to forecast occupant death risk is developed based on a parameter adjustment method of GridSearchCV, three algorithms with larger F1 values are selected from nine classification algorithms, which are voting classifier, gradient boosting, and decision tree. The results show that: ① travel direction, type of road sections, type of crushed vehicles, and different seats have significant effects on occupant death risk. The occupant death risk increases by 72% when compared accidents including vehicles driving in opposite directions with that including vehicles driving in the same direction. The risk decreases by 69% when compared accidents occur at non-freeway intersections with that occur at freeway sections. The occupant death risks for commercial trucks and commercial buses are 5 times and 3 times higher than passenger cars, respectively. The risk rises to around 8 times and 15 times for non-freeway non-intersection sections and freeway sections, respectively. The death risk for passenger seats increased by 70% compared with that for driver seat, and the death risk for passenger seat is nearly 4 times higher than that for driver seat at freeway sections. ② Different vehicles and road-section types are the most important features affecting occupant death. ③ The proposed model indicates that if a commercial truck collides at the front or rear of a 5-seater car at freeway sections or non-freeway non-intersection sections, the occupants have higher risk of death than chance of survival.
-
Key words:
- traffic safety /
- 5-seater cars /
- risk analysis /
- machine learning /
- feature crosses
-
表 1 变量说明
Table 1. Description of variables
变量 分类 必要说明 天气情况 无雨 事故发生为无雨天气 小雨 事故发生为小雨天气 中及大雨 事故发生为中、大雨天气 路段类型 高速 碰撞发生在高速公路路段 交叉 碰撞发生在非高速交叉路段 非交 碰撞发生在非高速非交叉路段 行驶方向 同向 目标车辆与碰撞车辆的行驶方向相同 异向 目标车辆与碰撞车辆的行驶方向不同 碰撞接触面 前面 目标车辆碰撞接触面在车头区域 侧面 目标车辆碰撞接触面在车头与车尾间 后面 目标车辆碰撞接触面在车尾区域 碰撞对象 乘用车 目标车辆与乘用车(包括皮卡车)的碰撞 商用客车 目标车辆与商用客车的碰撞事故 商用货车 目标车辆与商用货车发生碰撞 乘坐位置 前左 目标车辆前排左边的位置（驾驶位） 前右 目标车辆车前排右边的位置(副驾位） 后左 目标车辆后排左边的位置 后中 目标车辆后排中间的位置 后右 目标车辆后排右边的位置 死亡情况 是 目标车辆中的乘员死亡 否 目标车辆中的乘员存活 
下载: 导出CSV
表 2 变量分布情况（N1=1 028）
Table 2. Distribution of variables (N1=1 028)
特征变量 分类 存活人数(占比/%) 死亡人数(占比/%) χ2 天气情况 无雨 503(64.6) 276(35.4) 5.618* 小雨 105(59.0) 73(41.0) 中、大雨 37(52.1) 34(47.9) 路段类型 高速 129(55.4) 104(44.6) 66.839*** 交叉 207(84.8) 37(15.2) 非交 309(56.1) 242(43.9) 行驶方向 同向 278(64.8) 151(35.2) 1.335 异向 367(61.3) 232(38.7) 碰撞接触面 前面 464(59.8) 312(40.2) 12.174** 侧面 96(70.1) 41(29.9) 后面 85(73.9) 30(26.1) 碰撞对象 乘用车 240(86.0) 39(14.0) 97.611*** 商用客车 121(63.0) 7(37.0) 商用货车 284(51.0) 273(49.0) 乘坐位置 前左 297(69.2) 13(30.8) 17.176** 前右 135(54.0) 115(46.0) 后左 87(61.3) 55(38.7) 后中 4(56.8) 32(43.2) 后右 84(63.2) 49(36.8) 注：***，**，*分别表示置信水平为99%、95%、90%下显著。
下载: 导出CSV
表 3 共线性判断
Table 3. Collinearity judgment
特征变量 Tol VIF 天气情况 0.958 1.044 路段类型 0.730 1.369 行驶方向 0.754 1.327 碰撞接触面 0.894 1.119 碰撞对象 0.808 1.237 乘坐位置 0.980 1.021
下载: 导出CSV
表 4 二项Logistic回归分析结果
Table 4. Results of binary logistic regression
特征变量 分类 OR(95% C.I.) p值 天气情况 无雨 1 0.495 小雨 0.978(0.679~1.408) 0.905 中、大雨 1.37(0.801 ~2.346) 0.250 路段类型 高速 1 0.000 交叉 0.31(0.185 ~0.522) 0.000 非交 0.816(0.542~1.229) 0.331 行驶方向 同向 1 异向 1.719(1.190 ~2.482) 0.004 碰撞接触面 前面 1 0.209 侧面 0.709(0.461 ~1.089) 0.116 后面 0.774(0.465~1.287) 0.323 碰撞对象 乘用车 1 0.000 商用客车 3.126(1.950~5.013) 0.000 商用货车 4.797(3.178~7.239) 0.000 乘坐位置 前左 1 0.045 前右 1.703(1.202~2.415) 0.003 后左 1.129(0.739~1.723) 0.575 后中 1.310(0.766~2.240) 0.324 后右 1.065(0.688~1.648) 0.779 路段类型 & 碰撞对象 高速 & 乘用车 1 0.000 高速 & 商用客车 16.466(2.019~134.314) 0.009 非交 & 商用客车 7.950(1.251 ~50.508) 0.028 高速 & 商用货车 14.326(3.192~64.606) 0.001 非交 & 商用货车 8.169(1.253~53.256) 0.028 路段类型 & 碰撞接触面 高速 & 前面 1 0.000 高速 & 侧面 0.135(0.032~0.565) 0.006 交叉 & 侧面 9.219(1.871 ~45.433) 0.006 路段类型 & 乘坐位置 高速 & 前左 1 0.035 高速 & 前右 3.853(1.798~8.254) 0.001 注：OR值对应为1的为参照类别。
下载: 导出CSV
表 5 调参方法及结果
Table 5. Parameter adjustment methods and results
算法 参数 方法 取值 KNN n_neighbors n_neighbors= 3 DT max_depth（最大深度）min_samples_split（最小样本数）
min_samples_leaf（最少样本数）max_features（最大特征数）max_depth= 8 min_samples_leaf = 1
min_samples_split= 4 max_features= 1GB n_estimators（迭代次数）max_depth（最大深度）
min_samples_split（最小样本数）min_samples_leaf（最少样本数）max_features（最大特征数）learning_rate（步长）逐个调整直至F1取得最大值 n_estimators=323 max_depth=3
min_samples_split=5 min_samples_leaf=1
max_features=6 learning_rate= 0.1RF n_estimators（迭代次数）max_depth（最大深度）
min_samples_split（最小样本数）min_samples_leaf（最少样本数）max_features（最大特征数）n_estimators= 20 max_depth = 12
min_samples_split=2 min_samples_leaf= 1 max_features= 2
下载: 导出CSV
表 6 相关算法的F1值
Table 6. F1 value of correlation algorithms
F1值类别 GB RF DT NB SVM KNN AdaBoost LR VC 默认参数的F1值 0.549 0.533 0.518 0.500 0.463 0.539 0.498 0.368 调参后的F1值 0.579 0.547 0.565 0.548 0.587
下载: 导出CSV
-
[1] 中国共产党中央委员会, 中华人民共和国国务院. 交通强国建设纲要[EB/OL]. (2019-09)[2022-05-18]. http://www.gov.cn/zhengce/2019-09/19/content_5431432.htm.The Communist Party of China Central Committee and the State Council. Outline for building a leading transportation nation[EB/OL]. (2019-09)[2022-05-18]. http://www.gov.cn/zhengce/2019-09/19/content_5431432.htm. (in Chinese) [2] World Health Organization. Road traffic injuries[EB/OL]. (2021-06)[2022-05-18]. https://www.who.int/news-room/fact-sheets/detail/road-traffic-injuries. [3] 国家统计局. 中国统计年鉴(2017-2021)[M]. 北京: 中国统计出版社.National Bureau of Statistics of China. China statistical yearbook(2017-2021)[M]. Beijing: China Statistics Press. (in Chinese) [4] 刘金鑫. 汽车侧面柱碰结构安全性及乘员损伤仿真研究[D]. 兰州: 兰州交通大学, 2016.LIU J X. Simulation of cars structure safety and passenger injury in pole side impact[D]. Lanzhou: Lanzhou Jiaotong University, 2016. (in Chinese) [5] 肖媛. 车-车碰撞事故中驾乘人员伤害特征研究[D]. 西安: 长安大学, 2012.XIAO Y. The research on the injury of driver and passenger in vehicle-vehicle collisions[D]. Xi'an: Chang'an University, 2012. (in Chinese) [6] 支野, 王大珊, 丛浩哲, 等. 道路交通事故数据深度挖掘技术与应用-以深圳市为例[J]. 城市交通, 2018, 16(3): 28-32+61. https://www.cnki.com.cn/Article/CJFDTOTAL-CSJT201803007.htmZHI Y, WANG D S, CONG H Z, et al. Road traffic accident data analyzing and its application: Example of Shenzhen[J]. Urban Transport of China, 2018, 16(3): 28-32+61. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-CSJT201803007.htm [7] 李艳, 王雪松, 王婷, 等. 基于有限混合模型的高速公路事故影响因素分析[J]. 交通信息与安全, 2020, 38(6): 102-112. doi: 10.3963/j.jssn.1674-4861.2020.06.014LI Y, WANG X S, WANG T, et al. An analysis of impacting factors of freeway safety based on finite mixture models[J]. Journal of Transport Information and Safety, 2020, 38(6): 102-112. (in Chinese) doi: 10.3963/j.jssn.1674-4861.2020.06.014 [8] 赵丹, 马社强, 张雨萌, 等. 农村公路交叉口交通事故特征关联性与风险因素分析[J]. 中国安全科学学报, 2020, 30 (7): 146-151. https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK202007023.htmZHAO D, MA S Q, ZHANG Y M, et al. Correlation and risk factors analysis of traffic crash at intersections on rural highways[J]. China Safety Science Journal, 2020, 30(7): 146-151. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK202007023.htm [9] 张斌. 轿车侧碰撞安全性及乘员损伤防护技术研究[D]. 长沙: 湖南大学, 2010.ZHANG B. A study of safety performance of passenger car and occupant injury protection technique in side impact[D]. Changsha: Hunan University, 2012. (in Chinese) [10] 马聪, 张生瑞, 马壮林, 等. 高速公路交通事故非线性负二项预测模型[J]. 中国公路学报, 2018, 31(11): 176-185. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL201811020.htmMA C, ZHANG S R, MA Z L, et al. Nonlinear negative binomial regression model of expressway traffic accident frequency prediction[J]. China Journal of Highway and Transport, 2018, 31(11): 176-185. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL201811020.htm [11] 杨济匡, 覃祯员, 王四文, 等. 轿车侧面柱碰撞结构响应与乘员损伤研究[J]. 湖南大学学报(自然科学版), 2011, 38 (1): 23-28. https://www.cnki.com.cn/Article/CJFDTOTAL-HNDX201101006.htmYANG J K, QIN Z Y, WANG S W, et al. Study of the structural response and occupant injury in side pole impact to a passenger Car[J]. Journal of Hunan University(Natural Sciences), 2011, 38(1): 23-28. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-HNDX201101006.htm [12] 钮嘉颖, 郭刚, 丁玲, 等. 基于大壁障侧碰和75°侧柱碰的车身耐撞性和乘员损伤研究[J]. 汽车科技, 2017(5): 77-84. https://www.cnki.com.cn/Article/CJFDTOTAL-QCKJ201705014.htmNIU J Y, GUO G, DING L, et al. The research on vehicle crashworthiness and occupant injury based on AEMDB side impact and 75°side pole impact[J]. Auto Sci-Tech, 2017(5): 77-84. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-QCKJ201705014.htm [13] 司俊德, 李建平, 张博强, 等. 大客车正面碰撞乘客约束系统仿真分析[J]. 中国安全科学学报, 2021, 31(3): 135-141. https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK202103020.htmSI J D, LI J P, ZHANG B Q, et al. Simulation analysis of passenger restraint system in coach frontal collision[J]. China Safety Science Journal, 2021, 31(3): 135-141. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK202103020.htm [14] 颜凌波, 丁宗阳, 曹立波, 等. 车车斜角碰撞工况下驾驶员损伤研究[J]. 湖南大学学报(自然科学版), 2016, 43(4): 59-66. https://www.cnki.com.cn/Article/CJFDTOTAL-HNDX201604009.htmYAN L B, DING Z Y, CAO L B, et al. Study on the driver injury in vehicle to vehicle oblique crashes[J]. Journal of Hunan University(Natural Sciences), 2016, 43(4): 59-66. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-HNDX201604009.htm [15] 张光辉. 轿车正面斜角碰撞下后排左侧乘员损伤研究[D]. 成都: 西华大学, 2018.ZHANG G H. A study on rear left occupant injury under the frontal angle collision of the car[D]. Chengdu: Xihua University, 2018. (in Chinese) [16] 冯杰. 轿车偏置碰撞前排驾乘人员损伤特征仿真研究[D]. 重庆: 重庆交通大学, 2015.FENG J. The front occupant's injury characteristic simulation research in car offset collision[D]. Chongqing: Chongqing Jiaotong University, 2015. (in Chinese) [17] BOSE D, CRANDALL J, FORMAN J, et al. Epidemiology of injuries sustained by rear-seat passengers in frontal motor vehicle crashes[J]. Journal of Transport & Health, 2017, (4): 132-139. http://www.onacademic.com/detail/journal_1000039675611810_e2d9.html [18] SHIMAMURA M, YAMAZAKI M, FUJITA G. Method to evaluate the effect of safety belt use by rear seat passengers on the injury severity of front seat occupants[J]. Accident Analysis & Prevention, 2005, 37(1): 5-17. [19] BILSTON L E, DU W, BROWN J. A matched-cohort analysis of belted front and rear seat occupants in newer and older model vehicles shows that gains in front occupant safety have outpaced gains for rear seat occupants[J]. Accident Analysis & Prevention, 2010, 42(6): 1974-1977. http://europepmc.org/abstract/MED/20728650 [20] DURBIN D R, JERMAKIAN J S, KALLAN M J, et al. Rear seat safety: Variation in protection by occupant, crash and vehicle characteristics[J]. Accident Analysis & Prevention, 2015, 80(7): 185-192. http://onlinepubs.trb.org/onlinepubs/webinars/durbinoct272016.pdf [21] MITCHELL R J, BAMBACH M R, TOSON B. Injury risk for matched front and rear seat car passengers by injury severity and crash type: An exploratory study[J]. Accident Analysis & Prevention, 2015, 82(9): 171-179. http://www.researchgate.net/profile/Rebecca_Mitchell2/publication/278790420_Injury_risk_for_matched_front_and_rear_seat_car_passengers_by_injury_severity_and_crash_type_An_exploratory_study/links/5600862d08aec948c4fa8fdd.pdf [22] ISAKSSON-HELLMAN I, WERNEKE J. Detailed description of bicycle and passenger car collisions based on insurance claims[J]. Safety Science, 2017(92): 330-337. http://www.onacademic.com/detail/journal_1000038849900410_bb9c.html [23] 褚端峰, 吴超仲, 李顺喜, 等. 基于Logistic回归的高速公路车-车碰撞事故深度分析[J]. 中国安全科学学报, 2014, 24 (3): 103-108. https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK201403020.htmCHU D F, WU C Z, LI S X, et al. In-depth analysis of vehicle-vehicle crash on freeways based on logistic regression[J]. China Safety Science Journal, 2014, 24(3): 103-108. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK201403020.htm [24] PROCHOWSKI L, ZUCHOWSKI A. Analysis of the influence of passenger position in a car on a risk of injuries during a car accident[J]. Eksploatacjai Niezawodnosc - Maintenance and Reliability, 2014, 16(3): 360-366. http://yadda.icm.edu.pl/baztech/element/bwmeta1.element.baztech-ea4f138f-99f7-4b8f-afea-f09c7b19a49a/c/prochowski_2014-03-02_en.pdf [25] BECK L F, KRESNOW M J, BERGEN G. Belief about seat belt use and seat belt wearing behavior among front and rear seat passengers in the United States[J]. Journal of Safety Research, 2019, 68(1): 81-88. http://www.onacademic.com/detail/journal_1000041590505099_207d.html [26] 梅诗冬. 中国重大道路交通事故文本数据分析应用[D]. 河南焦作: 河南理工大学, 2017.MEI S D. Text data analysis and application of severe road traffic accidents in china[D]. Jiaozuo, Henan: Henan Polytechnic University, 2017. (in Chinese) -
点击查看大图
图(5) / 表(6)
计量
- 文章访问数: 1087
- HTML全文浏览量: 580
- PDF下载量: 667
- 被引次数: 0