A Machine Learning Framework for Vehicle Road Safety Risk Prediction and Identification

In an era of rapidly evolving transportation networks, the complexity of ensuring road safety demands more than traditional reactive measures. This study introduces a data-driven machine learning framework designed to proactively predict vehicle accident risks using structured traffic and environmental datasets. By integrating advanced preprocessing techniques—including feature engineering, one-hot encoding, SMOTE-based balancing, and normalization—the system prepares data for high-accuracy modeling. Four machine learning architectures were explored, with the Random Forest model demonstrating the most effective performance, achieving 96% accuracy, precision, recall, and F1-score. These consistent results across all metrics affirm its capability to handle imbalanced, multi-class accident severity prediction with reliability. The model draws on a range of features such as lighting conditions, vehicle types, road class, and time of day to identify risk-prone scenarios. Its interpretable design and balanced performance make it a practical asset for policymakers and traffic planners aiming to implement real-time, preventive strategies. The research validates the value of AI-powered safety frameworks in transforming reactive safety protocols into predictive, automated systems capable of safeguarding modern transport ecosystems