Comparative Analysis of Machine Learning Models for Accident Severity Prediction

Samudrala Tarunika, P Daphine Joy, Todupuniri Akshara Reddy, Vishnu KS — Fri, 03 Jan 2025 00:00:00 +0000

Accidents pose two major concerns: road safety and public health. The primary objective of this study was to develop an accident severity detection system that leverages machine learning algorithms to analyze a variety of influential factors, enabling the prediction of accident severity levels. The supervised learning algorithms employed in this system include Decision Trees, Naive Bayes, Support Vector Machines (SVM), Random Forest, and Logistic Regression, all aimed at providing accurate severity predictions. Key features incorporated in the training and testing datasets encompass driver demographics such as age, gender, education level, and driving experience, along with road characteristics like lane configurations and medians, junction types, and road surface conditions. Environmental factors such as light and weather conditions are also considered, as they may contribute to accident occurrence. Furthermore, accident-specific details, including collision types and vehicle/pedestrian movement patterns, are analyzed to uncover relationships and patterns influencing accident severity. The system produces a severity prediction score with associated probability, facilitating real-time alerts and warnings for stakeholders. This predictive model holds potential for improving road safety by enabling authorities and individuals to proactively mitigate the risk of severe accidents, especially when integrated with road safety initiatives. The research demonstrates the practical application of machine learning in predictive analytics, contributing to public safety efforts and informed policy-making.

International Journal of Human Computations & Intelligence

Comparative Analysis of Machine Learning Models for Accident Severity Prediction