A methodological framework for road accident severity prediction for indian highways using machine learning models

Area of Science:

• Road safety
• Machine learning applications
• Traffic accident analysis

Background:

• Road traffic accidents pose a significant global threat, with India experiencing over 150,000 annual fatalities.
• Existing models struggle to accurately represent the complex interactions of accident risk factors.
• There is a need for advanced analytical frameworks to understand and mitigate accident severity.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting road accident severity in India.
• To enhance model interpretability using SHAP values for identifying critical influencing factors.
• To provide a data-driven framework for improving highway safety and informing policy.

Main Methods:

• Implementation of Random Forest and Gradient Boosting algorithms for severity prediction.
• Application of SHAP (SHapley Additive exPlanations) values to determine feature importance.
• Evaluation of model performance using standard metrics like accuracy, precision, recall, and F1-score.

Main Results:

• Vehicle type, accident location, and road conditions were identified as significant predictors of accident severity.
• The SHAP-enhanced models provided clear insights into the contribution of each factor.
• Model performance metrics demonstrated the effectiveness of the proposed framework.

Conclusions:

• The SHAP-enhanced machine learning approach offers a robust and interpretable method for predicting road accident severity.
• Findings provide actionable insights for developing targeted road safety interventions in India.
• This framework supports proactive safety measures and infrastructure improvements on highways.