Accident Severity Prediction using Machine Learning Techniques

DOI

10.2991/978-94-6463-866-0_59How to use a DOI?

Keywords

Accident Severity; Machine Learning; Categorical Label Encoding and Median Imputation Pipeline with Feature Pruning; Random Forest Classifier; Support Vector Machine; RBF NN; BP NN; Classification

Abstract

Road traffic accidents are a critical public safety issue globally, causing substantial loss of life, serious injuries, and economic consequences. Accurate prediction of accident severity can enhance preventive measures, emergency response, and policy formulation. This study proposes a machine learning framework to classify accident severity into three categories—Fatal, Serious, and Slight—using historical accident data. The Random Forest (RF) classifier was found to perform best among various models, such as Radial Basis Function Neural Networks, Multi-Layer Perceptron, and Linear Support Vector Machines, achieving an overall accuracy of 87%. A comprehensive data preprocessing pipeline was applied to ensure data quality and consistency before model training. This included median imputation for missing values, label encoding for categorical variables, and data normalization. Feature selection further refined the dataset by eliminating low-importance variables, resulting in the top 20 most impactful features. The preprocessing pipeline ensured reproducibility and minimized data leakage throughout the workflow. The performance of the models was validated with detailed classification reports and confusion matrices. The Random Forest model outperformed other models, particularly in predicting minority classes like ‘Fatal,’ where RBFNN and Linear SVM showed limited sensitivity. Based on these results, the Random Forest classifier was chosen as proposed solution due to its high predictive power, resistance to overfitting, and interpretability. Looking ahead, Phase 2 of the project will involve integrating additional features related to human behavior, environmental conditions, road characteristics, and vehicle-specific factors to better capture the complexity of accident causation. Additional techniques were showcased and the random forest model has been improved and a real-time SOS alert system will be introduced to automatically notify emergency services in severe accidents. These developments aim to evolve the proposed model into a comprehensive, intelligent accident management system for smart transportation systems.

Copyright

© 2025 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  - V. Praveen
AU  - Golda Dilip
PY  - 2025
DA  - 2025/10/31
TI  - Accident Severity Prediction using Machine Learning Techniques
BT  - Proceedings of the International Conference on Intelligent Systems and Digital Transformation (ICISD 2025)
PB  - Atlantis Press
SP  - 715
EP  - 727
SN  - 2589-4919
UR  - https://doi.org/10.2991/978-94-6463-866-0_59
DO  - 10.2991/978-94-6463-866-0_59
ID  - Praveen2025
ER  -