| Title |
Neighborhood Environmental Factors Affecting Pedestrian Crash Severity - Using Machine Learning and eXplainable AI |
| Authors |
도수현(Do, Suhyun) ; 박승훈(Park, Seunghoon) |
| DOI |
https://doi.org/10.38195/judik.2025.10.26.5.77 |
| Keywords |
보행자 교통사고 심각도; 근린환경; 머신러닝; 설명가능한 인공지 Pedestrian Crash Severity; Neighborhood Environment; Machine Learning; ; Explainable AI |
| Abstract |
Traffic crashes impose substantial socio-economic burdens, including increaseds healthcare expenditure, productivity losses, and the exacerbation of social inequality.sThis study aims to identify neighborhood environmental factors associated with thesseverity of pedestrian crashes through an empirical analysis using an explainablesmachine learning model based on XGBoost. The result revealed that more than halfsof the ten most important variables were related to the neighborhood environments,shighlighting the explanatory role of urban spatial and physical conditions in shapingspedestrian crash severity. In addition, older adults, female pedestrians, and nighttimescrashes were associated with higher severity levels, emphasizing the importance ofscontext-sensitive design interventions and policy interventions tailored to vulnerablesgroups. Areas with a higher number of subway exits and bus stops showed lowerscrash severity. And the proportions of residential and commercial land use weresalso associated with mitigating pedestrian crash impacts. These findings suggest thatsintegrated spatial strategies combining transit-oriented urban design, land-usesplanning, and road environment improvements can effectively enhance pedestrianssafety. This study provides quantitative, data-driven evidence of the link betweensneighborhood environments and pedestrian crash severity and offers practicalsimplications for urban design and planning strategies aimed at traffic safety. |