Research and Analysis of Traffic Accident Severity Prediction Based on Data Augmentation and Feature Interpretation

DOI

10.2991/978-94-6239-648-7_74How to use a DOI?

Keywords

Traffic accident severity prediction; Data augmentation; SMOTE; Interpretability; SHAP

Abstract

Predicting the severity of traffic accidents is crucial to traffic safety management and accident prevention. However, in the actual data, the number of minor accidents far exceeds that of serious accidents, resulting in deviations in most types of predictions. In order to alleviate the category imbalance in traffic accident data, this paper proposes an analytical framework that combines SMOTE data enhancement and SHAP feature interpretation. Based on a traffic accident data set, this paper compares the performance differences between the four models of logical regression, support vector machine, random forest and XGBoost before and after data enhancement. The results show that SMOTE has effectively improved the model’s ability to identify a few classes, and the F1 value of SVM has been increased from 0.31 to 0.39. SHAP analysis shows that after sampling, the importance of characteristics such as “number of casualties”, “lighting conditions” and “road conditions” has increased significantly, indicating that the model pays more attention to key risk factors after data balance. This study validates the effectiveness of combining data augmentation and feature interpretation, providing a reference for traffic accident risk assessment.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Yuhan Chen
PY  - 2026
DA  - 2026/04/24
TI  - Research and Analysis of Traffic Accident Severity Prediction Based on Data Augmentation and Feature Interpretation
BT  - Proceedings of the International Workshop on Advances in Deep Learning for Image Analysis and Computer Vision (IWADIC 2025)
PB  - Atlantis Press
SP  - 685
EP  - 692
SN  - 2352-538X
UR  - https://doi.org/10.2991/978-94-6239-648-7_74
DO  - 10.2991/978-94-6239-648-7_74
ID  - Chen2026
ER  -