Development of a Model for Predicting the Severity of Traffic Accidents and Analysis of Accident Factors based on Machine Learning 


Vol. 14,  No. 2, pp. 72-81, Feb.  2025
https://doi.org/10.3745/TKIPS.2025.14.2.72


PDF
  Abstract

Traffic accidents pose significant societal costs and long-term impacts on victims, making research on prevention and severity reduction crucial. Accurate prediction of accident severity classes can provide valuable information for emergency response and reduce accident handling costs. Previous research on traffic accident severity prediction has been limited by its focus on single geographic areas and relatively low prediction accuracy. This study analyzed data from the Korea Expressway Corporation spanning 2022-2023, performing initial exploratory data analysis (EDA) and addressing multicollinearity through Variance Inflation Factor (VIF) analysis. Independent variables were normalized using Min-Max scaling and One-Hot encoding. The study compared the performance of multiple models: XGBoost, K-Nearest Neighbors (KNN), Random Forest, Support Vector Machine (SVM), Fully Connected Neural Network (FCNN), and Deep Neural Network (DNN). Random Forest emerged as the best-performing model, achieving 79.13% accuracy for the nationwide model through 5-fold cross-validation and hyperparameter optimization. By identifying the relative importance of various accident factors affecting severity levels, this study proposes customized policies to contribute to traffic accident prevention and severity reduction.

  Statistics
  Cite this article

[IEEE Style]

K. J. Hyuk and C. N. Wook, "Development of a Model for Predicting the Severity of Traffic Accidents and Analysis of Accident Factors based on Machine Learning," The Transactions of the Korea Information Processing Society, vol. 14, no. 2, pp. 72-81, 2025. DOI: https://doi.org/10.3745/TKIPS.2025.14.2.72.

[ACM Style]

Kim Jung Hyuk and Cho Nam Wook. 2025. Development of a Model for Predicting the Severity of Traffic Accidents and Analysis of Accident Factors based on Machine Learning. The Transactions of the Korea Information Processing Society, 14, 2, (2025), 72-81. DOI: https://doi.org/10.3745/TKIPS.2025.14.2.72.

Forms

Search
(IN TITLE, AUTHOR, ABSTRACT,KEYWORDS)

Advanced Search ▼

POPULAR KEYWORDS
(TOP 10 KEYWORDS)

Recent Publications
(LAST 3 YEARS)

Old Journals

Indexing

All publications of TKIPS is indexed in DOI, EBSCO, Google Scholar, Crossref, and CrossCheck.

TKIPS is also selected as the Journal for Accreditation by NRF (National Research Foundation of Korea).

Related Journals