Predictive Modeling of Road Accident Hotspots Using Open Traffic and Demographic Data

Road traffic accidents remain a major global challenge, causing significant loss of human life and placing heavy burden on public health systems and national economies. Although substantial efforts have been made to improve road infrastructure and traffic regulations, accident rates continue to rise because of increased vehicle ownership, urban expansion and complex traffic dynamics. A critical characteristic of road accidents is their tendency to occur repeatedly at specific locations, commonly referred to as accident hotspots. Accurate and timely identification of these locations is essential for effective road-safety management. Conventional hotspot identification methods primarily rely on historical accident data and descriptive techniques. Although such methods are useful for understanding past accident patterns, they lack predictive capability and often ignore dynamic traffic conditions, exposure differences, and demographic influences. This research proposes a predictive modeling framework that integrates open traffic data, demographic characteristics, and historical accident records to identify and forecast road accident hotspots at the road-segment level. Traffic-related indicators such as speed variability, congestion intensity, and temporal flow patterns are extracted from open traffic datasets, while demographic attributes provide insight into population exposure and land-use context. Machine learning models were trained to estimate accident risk and generate ranked hotspot lists. The experimental results show that the integration of traffic and demographic features significantly improves the prediction accuracy compared with traditional density-based approaches. The proposed system offers interpretable outputs through GIS-based visualization and an interactive dashboard, supporting proactive and data-driven road safety planning.