Prediction of Factors Related to Driver’s Behaviour in Occurring Road Traffic Accidents using Machine learning

Abstract

Road traffic accidents claim millions of lives annually and impose significant economic and societal costs. The relationship between key contributing factors and road traffic violations has been underexplored, while previous studies primarily focused on accident prediction. This study examined the impact of traffic violations across the top eight urban and rural states in the United States, analyzing the main causing factors, which were assessed using the Random Forest Machine Learning Model. Performance metrics, including accuracy, F-measure, precision, and Kappa statistics, validated the model’s effectiveness. Partial dependence plots explored the relationships between violations and contributing factors. In rural settings, violations were driven mainly by speeding, negligent driving, overcorrecting, and non-owner drivers. In urban areas, reckless driving, drug use, improper lane usage, tailgating, and failure to yield were the predominant factors. These findings underscore the need for tailored interventions to address area-specific violations, helping policymakers implement strategies to reduce violations and improve road safety.