Analysis of Influencing Factors for Nighttime Pedestrian-vehicle Crash Injury Severity Considering Temporal Instability
-
摘要: 夜间行人-机动车事故因能见度受限等因素导致伤害严重性显著高于白天。为精准识别其影响因素，构建1种混合方法，融合考虑均值和方差异质性的随机参数Logit模型与和基于沙普利可加性特征解释方法（SHapley Additive exPlanation，SHAP）的随机森林（random forest，RF）算法RF-SHAP，以2017—2022年的相关事故数据为研究对象，运用对数似然比检验对事故数据的时间稳定性进行评估，结果表明事故数据存在显著的时间不稳定性。为避免有偏的参数估计，按照2017—2019、2020、2021和2022年分别单独建模并计算显著变量平均边际效应。结果表明：①行人饮酒（2017—2019年）、救护车救援（2020年）、地方公路事故（2021年）及限速48~56 km/h（2022年）在对应年份具有随机效应，其均值或方差受交通控制、道路等级等变量影响；②行人饮酒、行人年龄＞45~60岁、驾驶员受伤、车辆类型为皮卡车、货车、道路双向有分隔、不同限速值（32~40 km/h、48~56 km/h、64~72 km/h）、周末和冬季近年来对夜间行人-机动车事故开始呈现显著影响。此外，借助RF-SHAP算法对模型中的随机参数变量进行特征贡献度分析，结果揭示了4个随机参数变量的所有子变量对事故严重程度的异质性影响，并提示在制定交通安全政策时，应重点关注行人饮酒问题，加强对高速与干线公路夜间事故的防控，并合理制定限速值，避免限速过高或过低。Abstract: Nighttime pedestrian-vehicle crashes exhibit significantly higher injury severity than daytime crashes due to visibility limitations and other factors. To accurately identify influencing factors, this study develops a hybrid model integrating a random parameters Logit model with heterogeneity in means and variances and a random forest (RF) algorithm based on SHapley Additive exPlanation (SHAP), i.e. RF-SHAP, using crash data from 2017 to 2022. The log-likelihood ratio test confirms temporal instability in the dataset, necessitating separate models for 2017—2019, 2020, 2021, and 2022 with calculated average marginal effects for significant variables. Results demonstrate that random effects exist for drinking pedestrians (2017—2019), ambulance required (2020), local street crashes (2021), and 48—56 km/h speed limits (2022), with their mean/variance influenced by traffic control and road classification. Drinking pedestrians, pedestrians aged over 45 to 60 years, driver injuries, vehicle types (pickup trucks and trucks), divided roadways, speed limits (32—40, 48—56, 64—72 km/h), weekends, and winter conditions have begun to exhibit statistically significant effects on nighttime pedestrian-vehicle crashes in recent years. In addition, the RF-SHAP algorithm quantifies heterogeneous contributions of all sub-variables within four random parameters to crash severity. Policy implications highlight three priorities: addressing pedestrian drinking behavior, enhancing nighttime crash prevention on expressways and arterial routes, and establishing appropriate speed limits while avoiding excessively high or low values.
-
表 1 变量描述表
Table 1. Variable description table
变量类别 变量 变量赋值 因变量 事故严重程度 0为无/可能伤害[30.9%]；1为轻伤（非致残伤害）[36.0%]；2为重伤（死亡/致残伤害）[32.9%] 行人特征 需救护车救援 0为否*[23.9%]；1为是[76.0%] 行人年龄/岁 0为≤25*[25.6%]；1为＞25~45[41.5%]；2为＞45~60[22.1%]；3为＞60[10.6%] 行人饮酒 0为否*[79.1%]；1为是[20.8%] 行人受伤 0为否*[4.0%]；1为是[95.9%] 行人性别 0为女性*[30.9%]；1为男性[69.0%] 驾驶人特征 驾驶人年龄/岁 0为≤25*[23.4%]；1为＞25~45[39.8%]；2为＞45~60[22.2%]；3为＞60[14.4%] 驾驶人饮酒 0为否*[95.2%]；1为是[4.7%] 驾驶人受伤 0为否*[94.0%]；1为是[5.9%] 驾驶人性别 0为女性*[35.1%]；1为男性[64.8%] 肇事逃逸 0为否*[73.2%]；1为是[26.7%] 碰撞特征 行车方向 0为非直行*[10.6%]；1为直行[89.3%] 车辆类型 0为小汽车*[55.6%]；1为SUV（运动型多用途汽车）[20.2%]；2为皮卡车[14.3%]；3为货车[6.7%]；4为其他[2.9%] 行人位于行车道 0为否*[27.6%]；1为是[72.3%] 行人穿过道路时发生碰撞 0为否*[62.6%]；1为是[37.3%] 道路特征 碰撞位置 0为交叉口及附近15m内*[35.5%]；1为非交叉口路段[64.4%] 道路线形 0为非弯曲*[94.5%]；1为弯曲[5.4%] 路面条件 0为干燥*[80.5%]；1为潮湿[18.9%]；2为冰霜雪冻[0.4%] 道路分隔情况 0为单向无分隔*[2.6%]；1为双向无分隔[65.1%]；2为双向有分隔[32.2%] 道路等级 0为高速公路*[3.9%]；1为干线公路[22.2%]；2为次干路[15.1%]；3为地方公路[58.5%] 道路限速/（km/h） 0为≤24*[1.6%]；1为32~40[10.9%]；2为48~56[34.9%]；3为64~72[30.1%]；4为≥80[22.2%] 交通控制 0为否*[60.0%]；1为是（交通标志及信号）[39.9%] 车道数/条 0为≤2*[51.8%]；1为＞2~4[28.5%]；2为＞4[19.5%] 时间和环境特征 不良天气 0为否*[75.4%]；1为是（雨/雪/雾）[24.5%] 工作区 0为否*[98.4%]；1为是[1.5%] 季节 0为春天*[19.0%]；1为夏天[18.9%]；2为秋天[31.3%]；3为冬天[30.5%] 有路灯照明 0为否*[53.0%]；1为是[46.9%] 周末 0为否*[70.8%]；1为是[29.1%] 地形条件 0为沿海*[28.6%]；1为山区[8.5%]；2为山麓[62.7%] 发展程度 0为乡村*[30.6%]；1为城市[69.3%] 事故发生区 0为商业区*[45.3%]；1为农林牧业区[18.3%]；2为工业和机构区[2.0%]；3为居住区[34.2%] 注：*表示自变量的参照类别，[]内为变量的各类别比率。 
下载: 导出CSV
表 2 相邻年份似然比检验结果
Table 2. Results of likelihood ratio test for adjacent years
(ti, ti+1) LL(βtiti) LL(βti) LL(βti+1) $\chi_\Upsilon ^2$值 自由度 置信水平/% （2017，2018） -2 214.23 -1 000.94 -1 167.05 92.49 84 75.34 （2018，2019） -2 283.99 -1 167.05 -1 069.07 95.73 82 85.74 （2019，2020） -2 142.73 -1 069.07 -1 018.98 109.37 82 97.66 （2020，2021） -2 128.74 -1 018.98 -1 053.56 112.40 84 97.90 （2021，2022） -2 168.74 -1 053.56 -1 055.68 119.01 86 98.93
下载: 导出CSV
表 3 2017—2019年似然比检验结果
Table 3. The results of the likelihood ratio test from 2017 to 2019
(ti-ti+2) $\chi_\Upsilon ^2$值 自由度 置信水平/% （2017—2019） 184.45 166 84.46
下载: 导出CSV
表 4 初步年份分组似然比检验结果
Table 4. Preliminary likelihood ratio test results for yearly grouping
年份 （2017—2019） 2020 2021 2022 （2017—2019） 608.74（74）[99.99%] 240.21（64）[99.99%] 451.65（62）[99.99%] 2020 106.17（66）[99.87%] 109.20（64）[99.96%] 182.74（62）[99.99%] 2021 95.36（66）[98.95%] 183.08（74）[99.99%] 167.06（62）[99.99%] 2022 109.60（66）[99.94%] 202.71（74）[99.99%] 129.32（64）[99.99%] 注：()里为自由度，[]为置信水平。
下载: 导出CSV
表 5 参数估计结果
Table 5. Estimation results for parameters
变量 2017—2019年 2020年 2021年 2022年 参数估计 z值 参数估计 z值 参数估计 z值 参数估计 z值 轻伤 常数项 -2.131 -4.47 -2.682 -5.51 -1.473 -4.35 -1.776 -2.99 需救护车救援 1.072 4.26 行人年龄＞45~60岁 0.519 2.71 行人特征 行人年龄＞60岁 0.750 2.72 0.721 3.29 0.756 3.21 行人饮酒 0.505 1.93 0.549 2.95 行人饮酒标准差 1.614 2.02 驾驶人特征 驾驶人饮酒 1.615 4.32 0.819 2.21 1.492 3.22 行车方向为直行 1.304 3.04 2.055 4.38 1.847 3.37 车辆类型为SUV 0.715 3.55 碰撞特征 车辆类型为皮卡车 0.524 2.03 行人位于行车道 0.672 2.95 0.549 2.97 0.838 3.78 0.784 4.16 行人穿过道路时发生碰撞 0.509 2.86 0.504 2.38 0.511 3.08 车道数3~4条 0.546 2.74 0.627 3.70 车道数＞4条 0.589 2.52 0.891 4.52 0.863 3.71 路面潮湿 0.542 2.32 限速32~40 km/h -1.934 -4.26 -2.035 -3.37 -2.236 -6.3 道路特征 限速48~56 km/h -0.973 -4.39 -2.119 -4.17 限速48~56 km/h标准差 1.643 1.97 限速64~72 km/h -0.809 -4.82 地方公路 -1.022 -6.56 -5.259 -3.47 地方公路标准差 2.940 3.35 时间和环境特征 冬季 0.515 2.58 有路灯照明 -0.681 -3.6 行人饮酒：有交通管控 0.900 2.10 均值异质性 地方公路：行车方向为直行 3.110 2.52 限速48~56 km/h：干线公路 1.205 1.81 方差异质性 地方公路：SUV 0.563 2.03 限速48~56 km/h：行人年龄＞45~60岁 0.787 1.76 重伤 常数项 6.267 4.66 1.141 4.28 0.723 5.3 救护车救援 -0.705 -4.14 -2.011 -3.52 -0.789 -4.89 -1.477 -9.96 救护车救援标准差 2.829 2.24 行人特征 行人年龄＞45~60岁 -0.555 -3.00 行人年龄＞60岁 -0.647 -2.48 行人受伤 -5.511 -4.13 驾驶人特征 驾驶人受伤 -1.396 -3.03 -0.935 -2.28 行车方向为直行 -0.719 -2.95 车辆类型为皮卡车 0.491 2.03 碰撞特征 车辆类型为SUV 0.371 2.21 0.503 2.16 车辆类型为货车 0.621 2.09 车辆类型为其他 -2.506 -2.21 行人位于行车道 -0.429 -2.71 道路特征 道路双向有分隔 -0.421 -2.55 夏季 -0.411 -2.21 秋季 时间和环境特征 冬季 0.496 3.22 周末 -0.508 -3.01 发展程度为城市 0.796 5.59 均值异质性 需救护车救援：驾驶人饮酒 -2.566 -2.01 方差异质性 需救护车救援：行人位于行车道 -0.631 -1.93 AIC 2118.7 2204.0 2273.0 2299.7 模型收敛时的对数似然 -1 036.363 -1 085.979 -1 116.484 -1 131.856 McFadden R2 0.172 0.128 0.125 0.144 注：设置无/可能伤害事故为参考类别。
下载: 导出CSV
表 6 显著影响因素平均边际效应值
Table 6. Average marginal effects of significant influencing factors
变量 2017—2019年 2020年 2021年 2022年 无/可能伤害 轻伤 重伤 无/可能伤害 轻伤 重伤 无/可能伤害 轻伤 重伤 无/可能伤害 轻伤 重伤 行人特征 需救护车救援 0.046 0.032 -0.078 -0.020 0.147 -0.128 0.085 0.062 -0.147 行人饮酒 -0.016 0.029 -0.013 -0.011 0.019 -0.007 行人受伤 0.667 0.332 -0.999 行人年龄＞45~60岁 -0.012 0.020 -0.009 0.012 0.005 -0.018 行人年龄＞60岁 -0.007 0.012 -0.005 -0.010 0.016 -0.005 0.007 0.003 -0.010 -0.009 0.016 -0.006 驾驶人特征 驾驶人饮酒 -0.008 0.013 -0.005 -0.005 0.006 -0.001 -0.004 0.008 -0.003 驾驶人受伤 0.004 0.003 -0.006 0.003 0.002 -0.005 碰撞特征 行车方向为直行 -0.104 0.179 -0.075 -0.240 0.358 -0.118 0.071 0.037 -0.108 -0.185 0.306 -0.122 车辆类型为SUV -0.010 -0.006 0.015 -0.021 0.018 0.003 车辆类型为皮卡车 -0.012 0.006 0.006 车辆类型为货车 -0.004 -0.003 0.007 车辆类型为其他 0.001 0.001 -0.002 行人位于行车道 -0.014 0.099 -0.085 -0.052 0.077 -0.026 -0.054 0.089 -0.035 -0.064 0.106 -0.042 行人穿过道路时发生碰撞 -0.017 0.026 -0.009 -0.012 0.020 -0.008 -0.018 0.030 -0.012 道路特征 车道数3~4条 -0.013 0.023 -0.010 -0.022 0.033 -0.011 车道数＞4条 -0.010 0.018 -0.007 -0.022 0.035 -0.012 -0.015 0.024 -0.010 路面潮湿 -0.007 0.013 -0.005 地方公路 0.060 -0.090 0.031 0.016 -0.022 0.006 道路双向有分隔 0.013 0.006 -0.020 限速32~40km/h 0.006 -0.011 0.005 0.006 -0.010 0.004 0.010 -0.017 0.006 限速48~56km/h 0.023 -0.040 0.017 0.024 -0.039 0.015 限速64~72km/h 0.035 -0.058 0.023 时间和环境特征 发展程度为城市 -0.051 -0.033 0.084 周末 0.012 0.008 -0.020 夏季 0.009 0.004 -0.013 冬季 -0.012 0.020 -0.008 -0.016 -0.012 0.027 有路灯照明 0.021 -0.037 0.015
下载: 导出CSV
-
[1] World Health Organization. Global status report on road safety 2023[R]. Geneva: World Health Organization, 2023. [2] STAMATIADIS N, PSARIANOS B, APOSTOLERIS K, et al. Nighttime versus daytime horizontal curve design consistency: issues and concerns[J]. Journal of Transportation Engineering, Part A: Systems, 2020, 146(3): 04019080. doi: 10.1061/JTEPBS.0000316 [3] MOUDON A V, LIN L, JIAO J, et al. The risk of pedestrian injury and fatality in collisions with motor vehicles, a social ecological study of state routes and city streets in King County, Washington[J]. Accident Analysis & Prevention, 2011, 43 (1): 11-24. http://pubmed.ncbi.nlm.nih.gov/21094292/ [4] 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) [5] CHARM T, WANG H, ZUNIGA-GARCIA N, et al. Predicting crash occurrence at intersections in Texas: an opportunity for machine learning[J]. Transportation Planning and Technology, 2024, 47(8): 1184-1204. doi: 10.1080/03081060.2023.2177651 [6] MANNERING F. Temporal instability and the analysis of highway accident data[J]. Analytic Methods in Accident Research, 2018, 17: 1-13. doi: 10.1016/j.amar.2017.10.002 [7] MADUSHANI J P S S, SANDAMAL R M K, MEDDAGE D P P, et al. Evaluating expressway traffic crash severity by using logistic regression and explainable & supervised machine learning classifiers[J]. Transportation Engineering, 2023, 13: 100190. doi: 10.1016/j.treng.2023.100190 [8] SERANEEPRAKARN P, HUANG S, SHANKAR V, et al. Occupant injury severities in hybrid-vehicle involved crashes: a random parameters approach with heterogeneity in means and variances[J]. Analytic Methods in Accident Research, 2017, 15: 41-55. doi: 10.1016/j.amar.2017.05.003 [9] HOU Q, HUO X, LENG J, et al. A note on out-of-sample prediction, marginal effects computations, and temporal testing with random parameters crash-injury severity models[J]. Analytic Methods in Accident Research, 2022, 33: 100191. doi: 10.1016/j.amar.2021.100191 [10] 潘义勇, 吴静婷, 缪炫烨. 老年驾驶员事故严重程度影响因素时间不稳定性分析[J]. 吉林大学学报(工学版). 2024, 54 (10): 2819-2826.PAN Y Y, WU J T, MIAO X Y. Temporal instability analysis of factors affecting injury severities of elderly drivers[J]. Journal of Jilin University(Engineering and Technology Edition). 2024, 54(10): 2819-2826. (in Chinese) [11] 蔡凌霄, 周备, 张生瑞, 等. 考虑均值及方差异质性的外卖骑手闯红灯行为影响因素分析[J]. 交通信息与安全, 2024, 42(1): 59-66. doi: 10.3963/j.jssn.1674-4861.2024.01.007CAI L X, ZHOU B, ZHANG S R, et al. Factors affecting red-light running behaviors of takeaway delivery riders considering heterogeneity in the means and variances[J]. Journal of Trasport Information and Safety, 2024, 42(1): 59-66. (in Chinese) doi: 10.3963/j.jssn.1674-4861.2024.01.007 [12] 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: 100113. doi: 10.1016/j.amar.2020.100113 [13] 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 (3): 1-16. http://cee.eng.usf.edu/faculty/flm/CGN6933/Heterogeneity.pdf [14] BEHNOOD A, MANNERING F L. The temporal stability of factors affecting driver injury severities in single-vehicle crashes: some empirical evidence[J]. Analytic Methods in Accident Research, 2015, 8(4): 7-32. http://www.sciencedirect.com/science?_ob=ShoppingCartURL&_method=add&_eid=1-s2.0-S2213665715000421&originContentFamily=serial&_origin=article&_ts=1475505860&md5=ec6b668a2836cccedf5d4ca80abc0967 [15] ISLAM M, ALNAWMASI N, MANNERING F. Unobserved heterogeneity and temporal instability in the analysis of work-zone crash-injury severities[J]. Analytic Methods in Accident Research, 2020, 28: 100130. http://www.sciencedirect.com/science/article/pii/S2213665720300208 [16] ALNAWMASI N, ALI Y, YASMIN S. Exploring temporal instability effects on bicyclist injury severities determinants for intersection and non-intersection-related crashes[J]. Accident Analysis & Prevention, 2024, 194: 107339. http://www.ncbi.nlm.nih.gov/pubmed/37857092 [17] 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) [18] 沈金星, 刘坤, 于淼, 等. 考虑时间稳定性的公交事故严重性影响因素分析[J]. 交通运输工程与信息学报, 2022, 20 (1): 108-118.SHEN J X, LIU K, YU M, et al. Analysis of factors influencing bus crash severity considering temporal stability[J]. Journal of Transportation Engineering and Information, 2022, 20 (1): 108-118. (in Chinese) [19] LI Y, SONG L, FAN W(DAVID). Day-of-the-week variations and temporal instability of factors influencing pedestrian injury severity in pedestrian-vehicle crashes: a random parameters Logit approach with heterogeneity in means and variances[J]. Analytic Methods in Accident Research, 2021, 29: 100152. [20] YAN X, HE J, ZHANG C, et al. Temporal instability and age differences of determinants affecting injury severities in nighttime crashes[J]. Analytic Methods in Accident Research, 2023, 38: 100268. [21] WASHINGTON S, KARLAFTIS M G, MANNERING F, et al. Statistical and econometric methods for transportation data analysis[M]. 3rd ed. New York: Chapman and Hall/CRC, 2020. [22] MCFADDEN D, TRAIN K. Mixed MNL models for discrete response[J]. Journal of Applied Econometrics, 2000, 15(5): 447-470. http://fisher.osu.edu/~schroeder.9/AMIS900/McFadden2000.pdf [23] BEHNOOD A, MANNERING F. Determinants of bicyclist injury severities in bicycle-vehicle crashes: a random parameters approach with heterogeneity in means and variances[J]. Analytic Methods in Accident Research, 2017, 16: 35-47. http://www.xueshufan.com/publication/2750077433 [24] 焦朋朋, 李汝鉴, 王健宇, 等. 考虑潜在类别的老年行人交通事故严重程度致因分析[J]. 交通运输系统工程与信息, 2022, 22(5): 328-336.JIAO P P, LI R J, WANG J Y, et al. Causes analysis on severity of elderly pedestrian crashes considering latent classes[J]. Journal of Transportation Systems Engineering and Information Technology, 2022, 22(5): 328-336. (in Chinese) [25] MILTON J C, SHANKAR V N, MANNERING F L. Highway accident severities and the mixed Logit model: an exploratory empirical analysis[J]. Accident Analysis & Prevention, 2008, 40(1): 260-266. [26] SONG L, LI Y, FAN W, et al. Mixed Logit approach to analyzing pedestrian injury severity in pedestrian-vehicle crashes in North Carolina: considering time-of-day and day-of-week[J]. Traffic Injury Prevention, 2021, 22 (7), 524-529. [27] MALYSHKINA N V, MANNERING F. Effect of increases in speed limits on severities of injuries in accidents[J]. Transportation Research Record, 2008, 2083(1): 122-127 http://www.documbase.com/goto/30575788-fc6e3ea9c78d6483cf5b46f7977f489c/Effect-of-Increases-in-Speed-Limits-on-Severities.pdf [28] SAM E F, DANIELS S, BRIJS K, et al. Modelling public bus/minibus transport accident severity in Ghana[J]. Accident Analysis & Prevention, 2018, 119: 114-121. http://pubmed.ncbi.nlm.nih.gov/30016751/ -
点击查看大图
图(2) / 表(6)
计量
- 文章访问数: 18
- HTML全文浏览量: 9
- PDF下载量: 4
- 被引次数: 0