基于模型群的重型货车交通安全风险因素分析

柯星安; 赵丹; 王秋鸿; 胡越宁; 牛帅

doi:10.3963/j.jssn.1674-4861.2024.04.008

基于模型群的重型货车交通安全风险因素分析

doi: 10.3963/j.jssn.1674-4861.2024.04.008

柯星安^1,,
赵丹^1, ,,
王秋鸿²,
胡越宁¹,
牛帅¹

1.
中国人民公安大学交通管理学院北京 100038
2.
公安部道路交通安全研究中心北京 100062

基金项目:

国家重点研发计划项目 2023YFB4302703

公安部科学研究项目 2023JSM05

详细信息

作者简介:
柯星安(2000—)，硕士研究生. 研究方向：交通安全. E-mail：lqlxdwl@163.com

通讯作者:
赵丹(1983—)，博士，副教授. 研究方向：交通安全、智能交通. E-mail：zhaodan@ppsuc.edu.cn

中图分类号: U491.31
计量
- 文章访问数: 38
- HTML全文浏览量: 40
- PDF下载量: 3
- 被引次数: 0
出版历程
- 收稿日期: 2023-07-10
- 网络出版日期: 2024-11-25

An Analysis of Risk Factors of Traffic Safety for Heavy Trucks Based on Model Group

1.
School of Traffic Management, People's Public Security University of China, Beijing 100038, China
2.
Research Institute for Road Safety of Ministry of Public Security, Beijing 100062, China

摘要

摘要: 为了深入探究重型货车交通事故的风险因素及发生机理，基于中国某省2016—2021年重型货车交通事故数据，构建集成随机森林、Logistic回归、地理加权Logistic回归和贝叶斯网络模型的模型群，对风险因素的影响程度、空间异质性及其因果路径进行分析。结果显示：①重型货车行驶状态、碰撞形态等10个因素对风险存在显著影响，其中农村交通参与者、正面和侧面碰撞在不同模型中的影响程度有轻微差异，追尾碰撞的影响程度在地理加权Logistic回归模型中较贝叶斯网络模型更高。②重型货车右转、存在违法行为、涉及弱势道路使用者时极容易发生亡人事故，分别使风险增加了41.9%，39.3%和39.0%。③以碰撞形态作为中介变量，重型货车行驶状态、事故另一方交通方式和年龄这3类因素与亡人事故风险的因果路径分析表明：当重型货车与弱势道路使用者发生侧面碰撞时，亡人事故风险比发生刮擦且事故另一方为其他类型机动车提高64.4%，为重型货车交通事故典型危险场景；对方年龄为30岁及以下时，追尾碰撞概率较30~60岁以及60岁以上分别增加10.3%和26.1%。④具有空间异质性的风险因素中，正面碰撞的空间异质性强度最大，右转的空间异质性强度最小。结论表明：基于模型群的分析框架可得到重型货车交通安全风险显著影响因素，可验证因素在不同模型中影响程度的差异性及空间异质性。
- 交通安全 /
- 重型货车 /
- 风险因素 /
- 贝叶斯网络 /
- 地理加权Logistic回归 /
- 因果路径
Abstract: To explore the risk factors and occurrence mechanisms behind the traffic accidents of heavy trucks in-depth, a model group, which comprises random forests (RF), Logistic regression (LR), geographically weighted Logistic regression (GWLR), and Bayesian network (BN), is established based on the data of heavy truck accidents in a certain province from 2016 to 2021. This model group allows for examining the impact magnitude, spatial het-erogeneity, as well as the causal pathways leading to accidents of the risk factors. The results reveal that: ①The driv-ing status of heavy trucks, collision patterns, and other eight factors have significant impacts on risks. The impact of rural traffic participants, as well as frontal and side collisions, varies slightly across different models, while the im-pact of rear-end collisions is higher in GWLR compared to that in BN.②When the heavy trucks are engaged in right turns, illegal behaviors or vulnerable road users, the risk of fatal accidents increases by 39.0%, 41.9% and 39.3%, respectively.③With the factor collision patterns serving as a mediator, the causal pathway analysis for the risk of fatal accidents indicates: the factors side collisions and vulnerable road users contribute to the increase of the risk of fatal accident by 64.4% compared to the impact of the factor scrapes involving other vehicles, therefore can be marked as the typical hazardous scenario. Furthermore, when the other participant of heavy truck accidents involves drivers not older than 30 years, the probability of rear-end collisions increases by 10.3% and 26.1% com-pared to those involving drivers aged 30-60 and above 60 respectively. ④Among the risk factors exhibiting spatial heterogeneity, the factor frontal collision exhibits the highest intensity, whereas the factor right turn shows the least. In conclusion, the analytical framework based on the model group can be used to identify the significant risk factors for traffic safety of heavy trucks, and to verify the differences of the impact across models and the notable spatial heterogeneity for these factors.
- traffic safety /
- heavy trucks /
- risk factors /
- Bayesian networks /
- geographically weighted Logistic regres-sion /
- causal pathways

HTML全文

图 1 模型群协同分析框架

Figure 1. The synergy interpreting framework of model group

下载: 全尺寸图片幻灯片

图 2 特征重要性(排名前13)

Figure 2. Feature importance(top 13)

下载: 全尺寸图片幻灯片

图 3 准确率与特征个数的学习曲线

Figure 3. Learning curve between accuracy and number of features

下载: 全尺寸图片幻灯片

图 4 树型贝叶斯网络结构

Figure 4. The Bayesian network structure for TAN

下载: 全尺寸图片幻灯片

图 5 回归系数空间分布

Figure 5. Spatial distribution of regression coefficients

下载: 全尺寸图片幻灯片

表 1 自变量选取及编码

Table 1. Selection and encoding of independent variables

归类	自变量	分类	编码	归类	自变量	分类	编码
事故信息	年份^[11]	2016	0	道路信息	地形	平原	0
		2017	1		地形	山区	1
		2018	2		线形	平直	0
		2019	3		线形	弯坡	1
		2020	4		照明条件	白天	0
		2021	5			黄昏或黎明	1
	月份	春(3~5)	0			夜间有路灯	2
		夏(6~8)	1			夜间无路灯	3
		秋(9~11)	2		路表情况	干燥	0
		冬(12~2)	3		路表情况	潮湿	1
	天气	良好(晴阴)	0		路口路段类型	普通路段	0
	天气	不良(雨雪雾)	1			特殊路段	1
	能见度/m	≥50	0			交叉口	2
	能见度/m	＜50	1			匝道口	3
	碰撞形态	刮擦	0		中央隔离设施	无	0
		正面碰撞	1		中央隔离设施	有	1
		侧面碰撞	2		路侧防护设施	无	0
		追尾碰撞	3		路侧防护设施	有	1
重型货车驾驶人信息	年龄/岁	≥20~30	0		道路类型	其他	0
	年龄/岁	＞30~60	1			高速公路	1
	驾龄/年	＞0~4	0			城市道路	2
		＞4~7	1			一般公路	3
		＞7	2		物理隔离	无隔离	0
	户口	城市	0			中心隔离	1
	户口	农村	1			机非隔离	2
	违法行为	无	0			中心加机非	3
	违法行为	有	1	对方交通参与者信息	对方性别	男	0
	车辆安全状态	正常	0		对方性别	女	1
	车辆安全状态	非正常	1		对方年龄/岁	≤30	0
	行驶状态	直行	0			＞30~60	1
		左转	1			＞60	2
		右转	2		对方户口	城市	0
		静止	3		对方户口	农村	1
	碰撞角色	碰撞	0		对方交通方式	其他机动车	0
	碰撞角色	被碰撞	1		对方交通方式	弱势道路使用者(VRU)	1

下载: 导出CSV

表 2 Logistic回归模型群结果对比

Table 2. Comparison of LRs model group results

自变量	影响因素	LR模型		GWLR模型
自变量	影响因素	回归系数	显著性	平均系数	最小系数	最大系数	空间异质性
对方户口	农村	1.141	＜0.001^*	1.144	0.828	1.304	-0.296^**
碰撞角色	被碰撞	0.077	0.664	0.102	0.067	0.132	0.501
违法行为	有	1.649	＜0.001^*	1.635	1.613	1.650	0.289
对方交通方式	VRU	1.578	＜0.001^*	1.549	1.499	1.622	0.395
物理隔离	中心隔离	-0.480	0.437	-0.484	-0.642	-0.078	0.398
	机非隔离	0.275	0.167	0.233	0.096	0.407	-0.391
	中心加机非	-0.293	0.528	-0.265	-0.456	-0.129	0.293
年龄/岁	＞30~60	0.251	0.241	0.232	0.207	0.329	0.244
年龄/岁	＞60	1.499	＜0.001^*	1.531	1.344	1.644	0.398
照明条件	黄昏或黎明	-0.113	0.762	0.010	-0.305	0.105	-0.014
	夜间有路灯	-0.008	0.974	-0.024	-0.076	0.167	0.437
	夜间无路灯	0.118	0.599	0.128	0.038	0.225	-0.016
碰撞形态	正面碰撞	1.493	＜0.001^*	1.494	0.885	1.757	-1.226^**
	侧面碰撞	0.705	0.003^*	0.631	0.586	0.733	0.310
	追尾碰撞	1.501	＜0.001^*	1.525	1.329	1.666	-0.380^**
年份	2017	-0.193	0.519	-0.145	-0.590	-0.061	0.174
	2018	-0.041	0.895	-0.059	-0.463	0.129	-0.530
	2019	-0.108	0.733	-0.158	-0.333	-0.071	0.434
	2020	-0.932	0.004^*	-0.950	-1.324	-0.801	0.019
	2021	-1.602	＜0.001^*	-1.644	-1.751	-1.554	0.384
行驶状态	左转	0.681	0.077	0.765	0.176	1.042	-0.781
	右转	3.148	＜0.001^*	3.232	2.718	3.397	-0.044^**
	静止	-0.033	0.902	-0.079	-0.182	0.158	-0.029
AICc		867.829		866.837
注：“”为该因素显著；“*”为该显著的因素存在空间异质性。

下载: 导出CSV

表 3 贝叶斯网络组合模型指标对比

Table 3. Comparison of BN combination model indicators

评价指标	RF-BN	RF-LR-BN
准确率	0.735	0.755
精确率	0.788	0.808
召回率	0.794	0.808
F1分数	0.791	0.808
AUC值	0.795	0.807

下载: 导出CSV

表 4 模型群影响程度整合结果

Table 4. Integration results of model group impact degree

风险因素	BN		GWLR		排名差值
风险因素	影响程度/%	排名	平均系数	排名	排名差值
违法行为_有	39.3	2	1.635	2	0
行驶状态_右转	41.9	1	3.232	1	0
对方户口_农村	13.7	6	1.144	7	1
对方交通方式_VRU	39.0	3	1.549	3	0
对方年龄(＞60)	29.2	4	1.531	4	0
碰撞形态_正面碰撞	17.1	5	1.494	6	1
碰撞形态_侧面碰撞	11.7	7	0.631	8	1
碰撞形态_追尾碰撞	4.2	8	1.525	5	3
年份(2020)	-21.1	9	-0.950	9	0
年份(2021)	-23.0	10	-1.644	10	0
注：当排名极差为0时，认为模型结果一致；当排名极差≠0时，认为存在一定程度的偏差。

下载: 导出CSV

表 5 相关因素对碰撞形态的影响程度

Table 5. The degree of influence on collision forms

影响因素	正面/%	侧面/%	追尾/%
行驶状态_左转	-1.8	32.0	-22.6
行驶状态_右转	-3.9	18.6	-23.0
行驶状态_静止	-7.3	-23.1	36.5
对方交通方式_VRU	0.2	23.9	-29.0
对方年龄_＞30~60	-4.8	8.2	-10.3
对方年龄_＞60	-4.7	16.8	-26.1

下载: 导出CSV

参考文献(26)

[1]	张洁, 张萌萌, 李虹燕. 基于二元Logistic模型的城市道路交通事故严重程度分析[J]. 交通信息与安全, 2022, 40(5): 70-79. doi: 10.3963/j.jssn.1674-4861.2022.05.008 ZHANG J, ZHANG M M, LI H Y. An analysis of severity of traffic accidents on urban roadways based on binary Logistic models[J]. Journal of Transport Information and Safety, 2022, 40(5): 70-79. (in Chinese) doi: 10.3963/j.jssn.1674-4861.2022.05.008
[2]	RAMÍREZ A F, VALENCIA C. Spatiotemporal correlation study of traffic accidents with fatalities and injuries in Bogota (Colombia)[J]. Accident Analysis & Prevention, 2021, 149: 105848.
[3]	朱彤, 秦丹, 魏雯, 等. 基于机器学习的公交驾驶员事故风险识别及影响因素研究[J]. 中国安全科学学报, 2023, 33 (2): 23-30. ZHU T, QIN D, WEI W, et al. Research on accident risk identification and influencing factors of bus drivers based on machine learning[J]. China Safety Science Journal, 2023, 33(2): 23-30. (in Chinese)
[4]	MA Z J, MEI G, CUOMO S. An analytic framework using deep learning for prediction of traffic accident injury severity based on contributing factors[J]. Accident Analysis & Prevention, 2021, 160: 106322.
[5]	MANNERING F, BHAT C R, SHANKAR V, et al. Big data, traditional data and the tradeoffs between prediction and causality in highway-safety analysis[J]. Analytic Methods in Accident Research, 2020, 25(C): 100113.
[6]	胡立伟, 吕一帆, 赵雪亭, 等. 基于数据驱动的交通事故伤害程度影响因素及其耦合关系研究[J]. 交通运输系统工程与信息, 2022, 22(5): 117-124, 134. HU L W, LYU Y F, ZHAO X T, et al. Influence factors and coupling relationship of traffic accident injury degree based on a data-driven approach[J]. Journal of Transportation Systems Engineering and Information Technology, 2022, 22(5): 117-124, 134. (in Chinese)
[7]	SUN Z Y, XING Y X, WANG J Y, et al. Exploring injury severity of vulnerable road user involved crashes across seasons: a hybrid method integrating random parameter logit model and Bayesian network[J]. Safety Science, 2022, 150: 105682. doi: 10.1016/j.ssci.2022.105682
[8]	THEOFILATOS A. Incorporating real-time traffic and weather data to explore road accident likelihood and severity in urban arterials[J]. Journal of Safety Research, 2017, 61: 9-21. doi: 10.1016/j.jsr.2017.02.003
[9]	AMINI M, BAGHERI A, DELEN D. Discovering injury severity risk factors in automobile crashes: a hybrid explainable AI framework for decision support[J]. Reliability Engineering & System Safety, 2022, 226: 108720.
[10]	SONG D D, YANG X B, YANG Y T, et al. Bivariate joint analysis of injury severity of drivers in truck-car crashes accommodating multilayer unobserved heterogeneity[J]. Accident Analysis & Prevention, 2023, 190: 107175.
[11]	周备, 孙晴, 张生瑞. 考虑时间不稳定性的货车事故严重程度分析[J]. 中国安全科学学报, 2022, 32(11): 160-167. ZHOU B, SUN Q, ZHANG S R. Severity analysis of freight car accidents considering time instability[J]. China Safety Science Journal, 2022, 32(11): 160-167. (in Chinese)
[12]	吕能超, 王玉刚, 周颖, 等. 道路交通安全分析与评价方法综述[J]. 中国公路学报, 2023, 36(4): 183-201. LYU N C, WANG Y G, ZHOU Y, et al. Review of road traffic safety analysis and evaluation methods[J]. China Journal of Highway and Transport, 2023, 36(4): 183-201. (in Chinese)
[13]	SANTOS K, DIAS J P, AMADO C. A literature review of machine learning algorithms for crash injury severity prediction[J]. Journal of Safety Research, 2022, 80: 254-269.
[14]	傅贵, 杨晓雨, 刘卓栩, 等. 安全科学的学科基本问题研究[J]. 中国安全科学学报, 2021, 31(5): 18-24. FU G, YANG X Y, LIU Z X, et al. Studies on fundamentals of safety science[J]. China Safety Science Journal, 2021, 31 (5): 18-24. (in Chinese)
[15]	戢晓峰, 詹换勤, 普永明, 等. 山区公路穿村镇路段过境车辆事故严重程度推理分析[J]. 交通运输系统工程与信息, 2022, 22(3): 231-237. JI X F, ZHAN H Q, PU Y M, et al. Inferential analysis of transit vehicle accident severity in mountain highway crossing villages and towns[J]. Journal of Transportation Systems Engineering and Information Technology, 2022, 22(3): 231-237. (in Chinese)
[16]	管理科学技术名词审定委员会. 管理科学技术名词[M]. 北京: 科学出版社, 2016. Approval Committee for the Chinese Terms in Management Science and Technology. Chinese terms in management science and technology[M]. Beijing: Science Press, 2016. (in Chinese)
[17]	马捷, 张云开, 蒲泓宇. 信息协同: 内涵、概念与研究进展[J]. 情报理论与实践, 2018, 41(11): 12-19. MA J, ZHANG Y K, PU H Y. Information collaboration: connotation, concept, and research progress[J]. Information Studies: Theory & Application, 2018, 41(11): 12-19. (in Chinese)
[18]	苏世亮, 李霖, 翁敏. 空间数据分析[M]. 北京: 科学出版社, 2019. SU S L, LI L, WENG M. Spatial data analysis[M]. Beijing: Science Press, 2016. (in Chinese)
[19]	YANG Z L, YANG Z S, SMITH J, et al. Risk analysis of bicycle accidents: a Bayesian approach[J]. Reliability Engineering & System Safety, 2021, 209: 107460.
[20]	宋述芳, 何入洋. 基于随机森林的重要性测度指标体系[J]. 国防科技大学学报, 2021, 43(2): 25-32. SONG S F, HE R Y. Importance measure index system based on random forest[J]. Journal of National University of Defense Technology, 2021, 43(2): 25-32. (in Chinese)
[21]	柯星安, 丁立民, 赵丹. 基于Logistic-TAN的电动自行车交通事故严重程度影响因素分析[J]. 中国人民公安大学学报(自然科学版), 2023, 29(2): 47-54. KE X A, DING L M, ZHAO D. Analysis on the influencing factors of traffic accidents severity for electric bicycles based on Logistic-TAN[J]. Journal of People's Public Security University of China(Science and Technology), 2023, 29(2): 47-54. (in Chinese)
[22]	SE C, CHAMPAHOM T, JOMMONKWAO S, et al. Modeling of single-vehicle and multi-vehicle truck-involved crashes injury severities: a comparative and temporal analysis in a developing country[J]. Accident Analysis & Prevention, 2024, 197: 107452.
[23]	杨硕. 重型货车交通事故严重程度影响因素及对策研究[D]. 北京: 中国人民公安大学, 2020. YANG S. Research on influence factors and counter measures of heavy goods vehicle traffic accident severity[D]. Beijing: People's Public Security University of China, 2020. (in Chinese)
[24]	BEHNOOD A, AL-BDAIRI N S S. Determinant of injury severities in large truck crashes: a weekly instability analysis[J]. Safety Science, 2020, 131: 104911.
[25]	陈强. 高级计量经济学及Stata应用[M]. 北京: 高等教育出版社, 2014. CHEN Q. Advanced econometrics and Stata applications[M]. Beijing: Higher Education Press, 2014. (in Chinese)
[26]	LIU J, HAINEN A, LI X B. Pedestrian injury severity in motor vehicle crashes: an integrated spatio-temporal modeling approach[J]. Accident Analysis & Prevention, 2019, 132: 105272.