摘要： 为探索建成环境对行人交通事故的影响并为行人事故预防提供理论依据，本文以建成环境“5D”要素为基础，围绕土地利用、城市设计和交通系统这3个维度构建昼夜建成环境指标体系，基于轻度梯度提升机构建昼-夜间行人交通事故严重程度模型，探究城市建成环境对行人交通事故严重程度的影响机制，结合SHAP(Shapley Additive Explanation)归因分析方法揭示两者之间的非线性关系，并以深圳市为例进行实证分析。结果表明：建成环境对行人交通事故的影响效应存在显著的时段异质性；昼间行人交通事故严重程度主要受人行道可达性、地铁站可达性及学校邻近度等因素的影响；夜间行人交通事故严重程度主要受人行道可达性、娱乐兴趣点(POI)指标及道路照明条件等因素的影响。建成环境对行人交通事故严重程度存在显著的非线性影响。昼间时段学校邻近度介于[0, 3]km时，地铁站可达性小于1km时，对行人事故严重程度有较大抬升作用；夜间时段娱乐POI可达性小于0.5km时，对行人事故严重程度有抬升作用；不论昼夜，人行道可达性对行人事故严重程度均有压降作用，且临街院门密度低的区域行人事故严重程度较 高。昼-夜间模型均表现出优秀的效果，分类准确率分别为96.38%和92.08%。

Abstract: The study on the impact of the built environment on pedestrian traffic accidents could provide a theoretical basis for accident prevention. This paper constructed a day-night built environment index system, based on three dimensions of land use, urban design, and transportation system in the "5D" elements. Using the light gradient boosting machine, a day-night pedestrian traffic accident severity model was constructed to explore the influence mechanism of the urban built environment on the severity of pedestrian traffic accidents. Combined with the SHAP attribution analysis method, the nonlinear relationship was revealed. Taking Shenzhen City as an example, the results show that there is significant temporal heterogeneity in the impact of the built environment on pedestrian traffic accidents. The severity of daytime pedestrian traffic accidents is mainly affected by factors such as sidewalk accessibility, subway station accessibility, and school proximity. At night, it is mainly affected by sidewalk accessibility, entertainment point of interest (POI) indicators, road lighting conditions, and other factors. The built environment has a conspicuous nonlinear effect on the severity of pedestrian traffic accidents. When the proximity of schools is between zero and three kilometers during the daytime and the accessibility of subway stations is less than three kilometers, it has a great effect on the severity of pedestrian accidents. When the accessibility of entertainment POI is less than 0.5 kilometers at night, it has a significant effect on the severity of pedestrian accidents. The accessibility of sidewalks can reduce the severity of pedestrian accidents both day and night, and the area with a low density of courtyard gates on the street has higherdegree accidents. Lastly, the model shows excellent results, with classification accuracies of 96.38% and 92.08%.

Key words: urban traffic, severity of pedestrian traffic accident, light gradient boosting machine (Light GBM), attribution analysis, built environment

戢晓峰, 乔新. 建成环境对行人交通事故严重程度的非线性影响[J]. 交通运输系统工程与信息, 2023, 23(1): 314-323.

JI Xiao-feng, QIAO Xin. Nonlinear Influence of Built Environment on Pedestrian Traffic Accident Severity[J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(1): 314-323.

[1] 何雅琴, 段雨阳, 王晨. 基于累积 Logistic 模型的行人交通事故严重程度分析及对策研究[J]. 安全与环境学报, 2021, 21(3): 1165-1172. [HE Y Q, DUAN Y Y, WANG C. Analysis of the pedestrian traffic accidents and countermeasures for heightening such safety based on the cumulative Logistic model[J]. Journal of Safety and Environment, 2021, 21(3): 1165-1172.]  [2] 陈坚, 刘柯良, 邸晶, 等. 建成环境对城市停车需求影响的非线性模型 [J]. 交通运输系统工程与信息, 2021, 21(4): 197-203. [CHEN J, LIU K L, DI J, et al. Nonlinear model of impact of built environment on urban parking demand[J]. Journal of Transportation Systems Engineering and Information Technology, 2021, 21(4): 197-203.]  [3] MEHRNAZ A, MEHMET B U, KARST T G, et al. A comprehensive analysis of the relationships between the built environment and traffic safety in the Dutch urban areas[J]. Accident Analysis & Prevention, 2022, 172(7): 1-17.  [4] MARTA R O, ALAN R S, CHRISTINE T N, et al. A new zone system to analyze the spatial relationships between the built environment and traffic safety[J]. Journal of Transport Geography, 2020, 84(4): 1-12.  [5] 戢晓峰, 张琪. 学区尺度下小学生通学事故风险评估及影响因素[J]. 交通运输系统工程与信息, 2021, 21 (1): 221-226. [JI X F, ZHANG Q. Risk assessment and influencing factors of pupils' school commuting accident risk in school district scale[J]. Journal of Transportation Systems Engineering and Information Technology, 2021, 21(1): 221-226.]  [6] RUI A, ZHAO M T, YI M D, et al. Examining non-linear built environment effects on injurious traffic collisions: A gradient boosting decision tree analysis[J]. Journal of Transport & Health, 2022, 24(3): 1-14.  [7] 袁振洲, 郭曼泽, 彭泳鑫, 等. 基于梯度关联规则的老年行人交通事故风险识别[J]. 交通运输系统工程与信息, 2022, 22(1): 195-208. [YUAN Z Z, GUO M Z, PENG Y X, et al. Risk recognition of older pedestrian traffic crashes based on XGB-Apriori algorithm[J]. Journal of Transportation Systems Engineering and Information Technology, 2022, 22(1): 195-208.]  [8] DING C, CHEN P, JIAO J F. Non-linear effects of the built environment on automobile-involved pedestrian crash frequency: A machine learning approach[J]. Accident Analysis & Prevention, 2018(112): 116-126.  [9] CHEN P, ZHOU J P. Effects of the built environment on automobile-involved pedestrian crash frequency and risk [J]. Journal of Transport & Health, 2016, 3(4): 448-456.  [10] ZAMANI A, BEHNOOD A, RASOUL S, et al. Temporal stability of pedestrian injury severity in pedestrianvehicle crashes: New insights from random parameter Logit model with heterogeneity in means and variances [J]. Analytic Methods in Accident Research, 2021, 32 (12): 1-29.  [11] CHANG I, PARK H, HONG E, et al. Predicting effects of built environment on fatal pedestrian accidents at location-specific level: Application of XG Boost and SHAP[J]. Accident Analysis & Prevention, 2022, 166(3): 1-11.  [12] 吴佩洁, 孟祥海, 曹梦迪. 城市交通事故多发点鉴别与时空模式挖掘[J]. 中国安全科学学报, 2020, 30 (11): 127-133. [WU P J, MENG X H, CAO M D. Identification of black spots in urban roads and spatiotemporal patterns mining[J]. China Safety Science Journal, 2020, 30(11): 127-133.]  [13] SERGIO G, SALVADOR G, JAVIER D S, et al. A practical tutorial on bagging and boosting based ensembles for machine learning: Algorithms, software tools, performance study, practical perspectives and opportunities[J]. Information Fusion, 2020 ,64(12): 205- [14] KJERSTI A, MARTIN J, ANDERS L. Explaining individual predictions when features are dependent: More accurate approximations to Shapley values[J]. Artificial Intelligence, 2021, 298(9): 1-23. 李文翔, 程佳楠, 刘向龙, 穆凯, 蔡近近, 刘魏巍. 出行即服务环境下个体出行链碳足迹监测与评估 [J]. 交通运输系统工程与信息, 2023, 23(2): 22-31. 金盛, 苏弘扬, 张静. 融合出行拓扑与序列分析的车辆时空出行模式挖掘 [J]. 交通运输系统工程与信息, 2023, 23(2): 40-53. 赵雪亭, 胡立伟. 基于典型耦合优化算法的城市交通拥塞点源反演识别研究 [J]. 交通运输系统工程与信息, 2023, 23(2): 74-83. 姚雅慧, 张戎. 基于轨迹数据的货车停车目的识别方法 [J]. 交通运输系统工程与信息, 2023, 23(2): 92-99. 李想, 晏启鹏, 骆晨. 空间异质性建成环境对地铁与公交换乘客流的影响 [J]. 交通运输系统工程与信息, 2023, 23(2): 100-110. 孙会君, 冯宗旭, 郑汉坤. 考虑地铁运营中断下多站协同的接驳公交优化研究 [J]. 交通运输系统工程与信息, 2023, 23(2): 111-120. 刘翔, 陈小鸿, 田茗舒. 成长性视角下建成环境对轨道交通站点客流影响分析 [J]. 交通运输系统工程与信息, 2023, 23(2): 121-127. 郑乐, 高良鹏, 陈学武. 基于多源数据的地铁公交换乘量影响因素与空间异质性分析 [J]. 交通运输系统工程与信息, 2023, 23(2): 128-138. 施俊庆, 李睿, 程明慧, 阮俊辉, 谢星. 基于动态时空神经网络模型的地铁客流预测 [J]. 交通运输系统工程与信息, 2023, 23(2): 139-147. 张兵, 周丹丹, 孙健, 倪训友. 基于双向长短期记忆网络的公交到站时间预测模型 [J]. 交通运输系统工程与信息, 2023, 23(2): 148-160. 李康康, 杨东峰. 影响老年人公交出行便捷性的个体与建成环境要素 [J]. 交通运输系统工程与信息, 2023, 23(2): 161-167. 潘寒川, 陆俊波, 胡华, 刘志钢, 沙悦. 客货混运下的城轨时刻表与流量控制协同优化研究 [J]. 交通运输系统工程与信息, 2023, 23(2): 187-196. 张惠茹, 豆飞, 魏运, 刘洁, 宁尧. 考虑列车运行净能耗与载客量差异的时刻表优化方法 [J]. 交通运输系统工程与信息, 2023, 23(2): 208-216. 黄赶祥, 张薇, 徐迪. 共享单车对我国主要城市交通拥堵的影响 [J]. 交通运输系统工程与信息, 2023, 23(2): 300-306.