A Hybrid Machine Learning-Based Real-Time Road Risk Prediction and Alert System

Road traffic accidents are a major global concern, causing significant loss of life, injuries, and economic damage. Identifying accident-prone zones and predicting potential risks are essential for improving road safety. This research presents a hybrid machine learning-based Real-Time Road Risk Prediction and Alert System designed to identify accident-prone areas and assess accident severity across Indian road networks. The proposed system integrates two machine learning algorithms: DBSCAN clustering for identifying accident hotspots and Random Forest classification for predicting accident severity. It also incorporates geospatial analysis, data preprocessing, dynamic risk scoring, and route risk assessment. The system utilizes an India Road Accident Dataset containing 3,000 accident records collected from 32 states and union territories between 2018 and 2023. OpenStreetMap geocoding and geospatial feature engineering are used to process location information and identify potential high-risk zones. The developed web-based application provides map-based visualization, route risk analysis, accident risk alerts, and alternative route suggestions to help users make safer travel decisions. The Random Forest model achieved an accuracy of 31.5%, highlighting the need for further improvement using larger and more accurate real-world datasets. This research demonstrates the potential of combining machine learning and geospatial technologies to support intelligent transportation systems and improve road safety. Keywords: Machine Learning, DBSCAN, Random Forest, Road Safety, Accident Prediction, Geospatial Analysis, Accident Hotspot Detection.