Traffic accident severity prediction based on an enhanced MSCPO-XGBoost hybrid model

Abstract Road traffic accidents pose a significant threat to public safety in China. This study proposes a novel severity prediction framework based on a Modified Stochastic Crested Porcupine Optimizer (MSCPO) combined with the XGBoost algorithm. The model was trained on 4287 accident cases from Chi...

Full description

Saved in:
Bibliographic Details
Main Authors: Fei Chen, Xiang Qun Liu, Jian Jun Yang, Xu Kang Liu, Jing Hui Ma, Jia Chen, Hua Yu Xiao
Format: Article
Language:English
Published: Nature Portfolio 2025-07-01
Series:Scientific Reports
Subjects:
Online Access:https://doi.org/10.1038/s41598-025-00797-7
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849387815470628864
author Fei Chen
Xiang Qun Liu
Jian Jun Yang
Xu Kang Liu
Jing Hui Ma
Jia Chen
Hua Yu Xiao
author_facet Fei Chen
Xiang Qun Liu
Jian Jun Yang
Xu Kang Liu
Jing Hui Ma
Jia Chen
Hua Yu Xiao
author_sort Fei Chen
collection DOAJ
description Abstract Road traffic accidents pose a significant threat to public safety in China. This study proposes a novel severity prediction framework based on a Modified Stochastic Crested Porcupine Optimizer (MSCPO) combined with the XGBoost algorithm. The model was trained on 4287 accident cases from China’s National Automobile Accident In-depth Investigation System (NAIS), collected between 2018 and 2023. The dataset was first divided into training and testing sets, and the Synthetic Minority Oversampling Technique (SMOTE) was applied only to the training set to address class imbalance. The MSCPO algorithm was then employed to optimize XGBoost hyperparameters. Comparative experiments demonstrate that the MSCPO-XGBoost model outperforms baseline algorithms including SVM, Random Forest, BP Neural Network, and CNN, achieving an accuracy of 83.57%, a recall of 85.23%, an F1-score of 84.30%, and an AUC of 92.82%. To enhance interpretability, SHAP analysis was used to identify key predictors such as engine displacement, vehicle mass, traffic signals, and driver age. The findings offer valuable guidance for traffic safety policymaking and demonstrate the potential of integrating real-time severity prediction into intelligent traffic management systems.
format Article
id doaj-art-f1f5f75d462642a2a3f2c46e18dfdcdd
institution Kabale University
issn 2045-2322
language English
publishDate 2025-07-01
publisher Nature Portfolio
record_format Article
series Scientific Reports
spelling doaj-art-f1f5f75d462642a2a3f2c46e18dfdcdd2025-08-20T03:42:29ZengNature PortfolioScientific Reports2045-23222025-07-0115111810.1038/s41598-025-00797-7Traffic accident severity prediction based on an enhanced MSCPO-XGBoost hybrid modelFei Chen0Xiang Qun Liu1Jian Jun Yang2Xu Kang Liu3Jing Hui Ma4Jia Chen5Hua Yu Xiao6School of Automobile and Transportation, Xihua UniversitySchool of Automobile and Transportation, Xihua UniversitySchool of Automobile and Transportation, Xihua UniversitySchool of Automobile and Transportation, Xihua UniversitySchool of Automobile and Transportation, Xihua UniversityChengdu Vocational & Technical College of IndustrySchool of Automobile and Transportation, Xihua UniversityAbstract Road traffic accidents pose a significant threat to public safety in China. This study proposes a novel severity prediction framework based on a Modified Stochastic Crested Porcupine Optimizer (MSCPO) combined with the XGBoost algorithm. The model was trained on 4287 accident cases from China’s National Automobile Accident In-depth Investigation System (NAIS), collected between 2018 and 2023. The dataset was first divided into training and testing sets, and the Synthetic Minority Oversampling Technique (SMOTE) was applied only to the training set to address class imbalance. The MSCPO algorithm was then employed to optimize XGBoost hyperparameters. Comparative experiments demonstrate that the MSCPO-XGBoost model outperforms baseline algorithms including SVM, Random Forest, BP Neural Network, and CNN, achieving an accuracy of 83.57%, a recall of 85.23%, an F1-score of 84.30%, and an AUC of 92.82%. To enhance interpretability, SHAP analysis was used to identify key predictors such as engine displacement, vehicle mass, traffic signals, and driver age. The findings offer valuable guidance for traffic safety policymaking and demonstrate the potential of integrating real-time severity prediction into intelligent traffic management systems.https://doi.org/10.1038/s41598-025-00797-7Accident severity predictionRoad traffic accidentXGBoostMSCPOSHAP
spellingShingle Fei Chen
Xiang Qun Liu
Jian Jun Yang
Xu Kang Liu
Jing Hui Ma
Jia Chen
Hua Yu Xiao
Traffic accident severity prediction based on an enhanced MSCPO-XGBoost hybrid model
Scientific Reports
Accident severity prediction
Road traffic accident
XGBoost
MSCPO
SHAP
title Traffic accident severity prediction based on an enhanced MSCPO-XGBoost hybrid model
title_full Traffic accident severity prediction based on an enhanced MSCPO-XGBoost hybrid model
title_fullStr Traffic accident severity prediction based on an enhanced MSCPO-XGBoost hybrid model
title_full_unstemmed Traffic accident severity prediction based on an enhanced MSCPO-XGBoost hybrid model
title_short Traffic accident severity prediction based on an enhanced MSCPO-XGBoost hybrid model
title_sort traffic accident severity prediction based on an enhanced mscpo xgboost hybrid model
topic Accident severity prediction
Road traffic accident
XGBoost
MSCPO
SHAP
url https://doi.org/10.1038/s41598-025-00797-7
work_keys_str_mv AT feichen trafficaccidentseveritypredictionbasedonanenhancedmscpoxgboosthybridmodel
AT xiangqunliu trafficaccidentseveritypredictionbasedonanenhancedmscpoxgboosthybridmodel
AT jianjunyang trafficaccidentseveritypredictionbasedonanenhancedmscpoxgboosthybridmodel
AT xukangliu trafficaccidentseveritypredictionbasedonanenhancedmscpoxgboosthybridmodel
AT jinghuima trafficaccidentseveritypredictionbasedonanenhancedmscpoxgboosthybridmodel
AT jiachen trafficaccidentseveritypredictionbasedonanenhancedmscpoxgboosthybridmodel
AT huayuxiao trafficaccidentseveritypredictionbasedonanenhancedmscpoxgboosthybridmodel