SafeRoad AI: Traffic Accident Risk Prediction Using Machine Learning Regression

SafeRoad AI is a machine-learning prototype that estimates traffic accident risk as a continuous score using Random Forest Regression. It uses a synthetic dataset of 5,000 traffic observations with nine inputs: traffic volume, average speed, visibility, vehicle density, time of day, weather, road condition, junction presence, and weekend status. Categorical variables are one-hot encoded, and a Random Forest Regressor (200 trees, maximum depth 12) is trained on an 80:20 train-test split.On the test set, the model achieved an MAE of about 1.258, an MSE of about 2.439, and an R² of about 0.730. Average speed, traffic volume, vehicle density, and visibility were the most influential features. The project also includes a Streamlit application with a dashboard, interactive risk prediction, model analytics, and future scope.Because the data and risk score are simulated, the results demonstrate a machine-learning workflow, not real-world accident probabilities. The paper also discusses limitations, responsible AI considerations, and future work such as real-time sensors, weather APIs, geographic risk mapping, and explainable AI.