AI DRIVEN ROAD ACCIDENT PREDICTION USING MACHINE LEARNING AND DEEP LEARNING FOR INDIAN ROAD CONDITIONS

Abstract

One of the causes of injury, deaths and loss of money in India is road accidents. The primary causes of high accident rates include urbanisation, traffic jams, weather variations, and high speed. Instead of preventing accidents, the traditional traffic systems mostly respond to them. This research presents a machine learning and deep learning-based smart road accident risk prediction system. A realistic Indian road dataset of 1000 records was created using important factors including time, weather, road type, traffic density, speed limit, visibility, existence of junction, and day of the week. The preprocessed dataset was used to train Random Forest and Artificial Neural Network models. According to experiments, the Random Forest model's accuracy on the testing data was 88.5%, while the Deep Learning model's accuracy was 81.99. High prediction capacity for low, medium, and high accident risk is confirmed by classification metrics and confusion matrices. The proposed system will be useful for proactive road safety management in smart cities, navigation systems, traffic departments, and emergency services.