Abstract: Road traffic accidents remain a critical public health challenge in India, accounting for over 150,000 deaths annually and 11% of global fatalities despite only 1% of the world's vehicles. Current traffic management systems are reactive, focusing on post-incident response rather than proactive prevention. This paper introduces the Vehicle Collision Analysis Engine (VCAE), a hybrid ensemble machine learning platform integrating Random Forest and XGBoost with geospatial analytics and explainable AI (XAI). Using a synthetic dataset aligned with Ministry of Road Transport and Highways (MoRTH) distributions, the system predicts accident severity, dentifies emerging 'Greyspots' before they escalate, and provides transparent, actionable Results recommendations. demonstrate an R^2 of 0.89, outperforming standalone models. Greyspot validation achieved 74% accuracy. User acceptance testing yielded 87% satisfaction. Deployment simulations confirmed sub-second response times for 500 users. Nationwide deployment suggests potential annual savings of 15,000+ lives. VCAE represents a replicable framework for proactive, explainable, and scalable traffic safety management in emerging economies.

Index Terms: Traffic Safety, Machine Learning, Explainable AI, Ensemble Models, Geospatial Analytics, Risk Prediction, Intelligent Transportation Systems.

Downloads: | DOI: 10.17148/IARJSET.2026.13315