基于XGBoost和混合Logit模型的机动车对撞事故受伤严重程度分析

打开文本图片集

关键词：交通安全；伤害严重程度；机器学习；XGBoost模型；混合Logit模型中图分类号：F540；U491.31 文献标志码：A DOI： 10.13714/j.cnki.1002-3100.2025.12.013

Abstract：Thisaticleanalyzestheinfluencingfactorsof head-oncolionsinvolvingmotorvehicles，inlightoftherisingnumberof suchacidentsThearticleobtainstraffcacidentdatafromNorthCarolina，USA，coveringtheyearsfrom2O13to2017. Itcategorizestheseverityofacidentsbasedontheinjuryseverityofpeopleinvolved，using fourmainaspects：humanfactors， vehicleharacteristics，roadconditionsandenvironmentalinfluences.Atotalof41factorsareselectedasindependentvariables， whiletheseverityofinjuriesservesasthedependentvariable，ategorizedintofivelevels：noinjuries，minorinjuries，serious non-disablinginjuries，seriousdisablinginjuries，andfatalinjuries.ApredictivemodelisestablishedbasedontheExtreme GradientBosting （XGBost）algorithmandmixedLogitmodel.First，theaccidentdataisanalyzedusingtheXGBostpredictive modeltoidentifythetop2Oindependentvariablesthatsignificantlyimpacttraficaccidents.Thesevariablesarethenputinto themixed Logit model to filter for independent variables with a significance level of P<0.05 .The results indicate that four variables-drivercharacteristcs（drivingudertheinfluence），roadcharacteristics（curvature）oadconditions（icysurfaces）and functionalareas （farmlands，forests，pastures）—exhibitrandomorparametercharacteristics.Ananalysisofthemarginaleffects from the mixed Logit model shows that when vehicle speed limits exceed 60km/h ，the risk of serious disabling injuriesis increased by 0.71% ，and the probability of fatal accidents rises by 0.83% ，compared to speed limits between 36km/h and 60km/h . Basedonthesefindings，measurescanbeakentopreventtheoccurenceofsuchaccidents，providingdatasupportandaheoretical basis for traffic managers and policymakers.

KeyWords：traffic safety;injuryseverity;machine learning;XGBoostmodel;mixedLogitmodel

0引言

世界卫生组织最新发布的《2023年道路安全全球现状报告》中指出，自2010年以来，道路交通死亡人数每年下降 5% 降低至每年119万人。（剩余14374字）