Ahmed Hossain, Xiaoduan Sun, A. S. Hasan, M. Jalayer, Julius Codjoe
{"title":"Comprehensive Investigation of Pedestrian Hit-and-Run Crashes: Applying XGBoost and Binary Logistic Regression Model","authors":"Ahmed Hossain, Xiaoduan Sun, A. S. Hasan, M. Jalayer, Julius Codjoe","doi":"10.1177/03611981241262315","DOIUrl":null,"url":null,"abstract":"The present trend in the United States suggests that one in five pedestrian fatalities in motor vehicle crashes involves a hit-and-run, a serious traffic safety concern. The over-representation of pedestrian hit-and-run collisions necessitates a systemic data-driven investigation to uncover the contributing factors that cause fatalities or serious injuries. This study addressed two research questions (RQ), RQ1: What factors contribute to pedestrian hit-and-runs? RQ2: What causes hit-and-run pedestrian fatalities? This study addresses the RQs using the XGBoost algorithm (RQ1) and binary logistic regression model (RQ2) to analyze police-reported pedestrian crashes (2015–2019) in Louisiana. The XGBoost model was used to classify pedestrian hit-and-run crashes (hit-and-run = yes/no) and identified critical factors as predictors of pedestrian hit-and-run crashes including: primary contributing factors (pedestrian action, pedestrian violation, prior movement, pedestrian condition); settings (dark-with-streetlight, posted speed limit of > 55 mph, two-way road with physical separation); pedestrian characteristics (younger and older pedestrians, male gender, presence of dark clothing); and weekend. The binary logistic regression model was further used to identify critical high-risk hit-and-run scenarios resulting in fatal or severe injury of pedestrians. Some of the identified top factors are posted speed limit of 55 mph or higher (OR = 12.74), pedestrian impairment (OR = 4.77), older pedestrians (OR = 2.68), younger pedestrians (OR = 1.79), and dark-no-streetlight conditions (OR = 2.91). Both models showed strong relationships between pedestrian hit-and-run crashes and fatal or severe injuries (e.g., dark-with-streetlight, high-speed settings, older pedestrians, and pedestrian actions). Identifying these critical links can help policymakers, law enforcement agencies, and transportation authorities develop targeted interventions and strategies to address the risk factors.","PeriodicalId":309251,"journal":{"name":"Transportation Research Record: Journal of the Transportation Research Board","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2024-08-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Transportation Research Record: Journal of the Transportation Research Board","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1177/03611981241262315","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
The present trend in the United States suggests that one in five pedestrian fatalities in motor vehicle crashes involves a hit-and-run, a serious traffic safety concern. The over-representation of pedestrian hit-and-run collisions necessitates a systemic data-driven investigation to uncover the contributing factors that cause fatalities or serious injuries. This study addressed two research questions (RQ), RQ1: What factors contribute to pedestrian hit-and-runs? RQ2: What causes hit-and-run pedestrian fatalities? This study addresses the RQs using the XGBoost algorithm (RQ1) and binary logistic regression model (RQ2) to analyze police-reported pedestrian crashes (2015–2019) in Louisiana. The XGBoost model was used to classify pedestrian hit-and-run crashes (hit-and-run = yes/no) and identified critical factors as predictors of pedestrian hit-and-run crashes including: primary contributing factors (pedestrian action, pedestrian violation, prior movement, pedestrian condition); settings (dark-with-streetlight, posted speed limit of > 55 mph, two-way road with physical separation); pedestrian characteristics (younger and older pedestrians, male gender, presence of dark clothing); and weekend. The binary logistic regression model was further used to identify critical high-risk hit-and-run scenarios resulting in fatal or severe injury of pedestrians. Some of the identified top factors are posted speed limit of 55 mph or higher (OR = 12.74), pedestrian impairment (OR = 4.77), older pedestrians (OR = 2.68), younger pedestrians (OR = 1.79), and dark-no-streetlight conditions (OR = 2.91). Both models showed strong relationships between pedestrian hit-and-run crashes and fatal or severe injuries (e.g., dark-with-streetlight, high-speed settings, older pedestrians, and pedestrian actions). Identifying these critical links can help policymakers, law enforcement agencies, and transportation authorities develop targeted interventions and strategies to address the risk factors.