AI-BASED ACCIDENT ZONE PREDICTION AND RISK MAPPING SYSTEM

Abstract

The escalating frequency of road accidents, often driven by unpredictable environmental variables, necessitates a shift from traditional retrospective mapping to intelligent, proactive safety technologies. Traditional safety systems, which rely heavily on static historical data, are increasingly insufficient for addressing the dynamic nature of modern road hazards. This project introduces a comprehensive AIBased Accident Zone Prediction and Risk Mapping System designed to mitigate these risks through Real-Time Contextual Analysis. The proposed framework utilizes a Hybrid Risk Scoring Algorithm as a data fusion engine, continuously integrating historical FIR records, live atmospheric conditions, and traffic velocity to identify high-risk zones in real-time. Leveraging robust ensemble learning techniques—specifically Random Forest and XGBoost—the system forecasts accident probabilities with high precision, effectively handling non-linear relationships between road attributes and crash frequency. A key innovation of this system is the integration of a Generative AI Reporting Module powered by Large Language Models (LLMs), which transforms complex analytics into automated, human-readable safety insights for authorities. Additionally, the platform features a SafeRoutemNavigation mechanism that proactively guides commuters away from identified critical danger zones based on live safety scores. By visualizing risks through dynamic heatmaps and interactive geospatial layers, this system bridges the gap between reactive administrative recordkeeping and proactive accident prevention, offering a scalable solution to significantly reduce road fatalities.