ROAD TRAFFIC ACCIDENT RISK PREDICTION AND KEY FACTOR IDENTIFICATION FRAMEWORK BASED ON EXPLAINABLE DEEP LEARNING

This paper presents a novel framework for predicting road traffic accident risk and identifying key contributing factors using explainable deep learning techniques.The system integrates a hybrid CNN-BiLSTM-Attention model to capture spatial dependencies from road environment features and temporal patterns from traffic flow data.To ensure interpretability, DeepSHAP is employed for feature attribution, enabling transparent identification of critical accident risk factors.The methodology emphasizes comprehensive dataset preprocessing, including traffic density normalization, weather condition encoding, and temporal sequence modeling, followed by deep learning model training and validation.Experimental evaluation on benchmark traffic datasets demonstrates that the proposed framework significantly outperforms conventional machine learning and baseline deep learning approaches in terms of accuracy, recall, and F1-score.The explainability layer highlights influential features such as vehicle volume, road geometry, speed variations, and weather conditions, thereby providing actionable insights for transportation authorities.The expected outcomes include a reliable decision-support system that not only predicts accident-prone scenarios in advance but also guides policymakers toward effective safety interventions.By combining predictive performance with interpretability, this framework enhances trust in AI-driven traffic management and contributes toward proactive accident prevention and improved road safety