Predicting the severity of traffic accidents is crucial for enabling timely and effective rescue measures. However, there is limited research that offers accurate predictions in this domain. Hence, in this paper, we employ data preprocessing and feature extraction on the US Accidents data set spanning from 2016 to 2023. We construct a LightGBM model to predict traffic accident severity. Additionally, we employ SHAP value analysis of the influence of various features on accident severity. Our study demonstrates that the LightGBM model we built is highly effective in predicting traffic accident severity, achieving a precision of 80.57%, a recall of 80.20%, and an F1-Score of 80.02%. Furthermore, we also found that the source, year, accident impact distance, month, and region had the greatest impact on the severity of the accident. These findings suggest that our research enhances the efficiency of emergency responses to traffic accidents and reduces their impact on the transportation system.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Research on Traffic Accident Severity Prediction Based on the LightGBM Model


    Contributors:
    Hao, Yanjun (author) / Dong, Hongxia (author) / Xu, Ting (author) / Zhang, Zhishun (author)

    Conference:

    24th COTA International Conference of Transportation Professionals ; 2024 ; Shenzhen, China


    Published in:

    CICTP 2024 ; 2529-2539


    Publication date :

    2024-12-11




    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English



    Traffic accident severity prediction based on interpretable deep learning model

    Pei, Yulong / Wen, Yuhang / Pan, Sheng | Taylor & Francis Verlag | 2025


    Using LightGBM with SHAP for predicting and analyzing traffic accidents severity

    Li, Jinqiang / Guo, Yuying / Li, Li et al. | IEEE | 2023


    Prediction of road traffic accident severity based on multi-model fusion

    Lu, Wenting / Huang, Mingxia / Yu, Rongze | British Library Conference Proceedings | 2023