Analysis of Road Traffic Accident Using AI Techniques

Abstract

Road traffic accidents are one of the global safety and socioeconomic challenges. According to WHO (2024), it has caused over 1.19 million annual fatalities. It is also projected to cause economic losses, which are approximately $1.8 trillion between 2015 and 2030. In this research, machine learning (ML) approach was implemented to predict the severity of road traffic accidents and explore actionable insights for intervention. The dataset used in implementing machine learning models was collected from Victoria Road Crash incidence from the years 2012-2023. This dataset includes temporal, environmental, and infrastructure variables. The target variable is severity of the road accident which is in four classes: fatal, serious injury, minor injury, and property damage. The first part of the machine learning analysis involves feature analysis using feature importance by random forest and partial dependence plots. The feature analysis identified temporal factors like accident time and date as key influencing factors of severity. The significant peaks from feature analysis showed rush hours and late weekdays as major determinants of road accidents in Victoria. Similarly, speed zones also showed a significant influence on road accidents, and this emphasizes the correlation between higher speed limits and severe outcomes. Environmental and infrastructural factors, like lighting conditions and road geometry, showed comparatively lower impact. In the second part of the analysis, three machine learning models—Logistic Regression, Random Forest, and XGBoost—were implemented for predictive performance. Logistic Regression outperformed others with the classification of minor injuries (Class 3), with a recall of 100%. Random Forest showed slightly better balance across classes. However, all models struggled with minority classes, like fatal accidents (Class 1), due to class imbalance. Overall, the findings revealed the importance of targeted interventions during high-risk periods with stricter speed limit enforcement and improved lighting infrastructure.