Severity Prediction of Jordan Road Accidents using Artificial Intelligence

Abstract

Road traffic accidents are a significant global concern, with developing countries accounting for 85% of annual fatalities and 90% of disability-adjusted life years lost. This study investigates the severity of road accidents in Jordan using a machine learning-based predictive approach. A dataset of 73,000+ accident reports from 2018 was analyzed, covering factors such as road conditions, weather, vehicle attributes, and driver demographics. The primary objective is to develop and evaluate machine learning models for predicting accident severity. Seven classification algorithms were tested: Logistic Regression (LR), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Naïve Bayes (NB), Random Forest (RF), Decision Tree (DT), and Extreme Gradient Boosting (XGBoost). The results indicate that LR achieved the highest accuracy at 98.1%, followed by RF (95.02%) and XGBoost (95.27%). Feature importance analysis revealed that road type, lighting conditions, and driver violations were the most influential factors in predicting accident severity. A key novelty of this research is the integration of real-world Jordanian accident data with machine learning models to enhance predictive accuracy. The study's findings provide actionable insights for policymakers, enabling targeted interventions to reduce accident severity. The dataset is made publicly available to support future research. This research contributes to the advancement of AI-driven traffic safety solutions, demonstrating the effectiveness of machine learning in real-time risk assessment and decision-making.