Momahmoses/road-accident-hotspot

KDE hotspot analysis and ML severity prediction to support FRSC road safety interventions across Nigeria. # Road Accident Hotspot Analysis & Traffic Safety GIS + spatial statistics system mapping road accident hotspots on Nigerian highways, supporting the Federal Road Safety Corps (FRSC) and state governments in targeting high-risk corridors for intervention. --- ## Problem Statement Nigeria records over 10,000 road fatalities annually. Without data-driven hotspot maps, FRSC resources are deployed reactively. This system identifies where accidents cluster, why they happen, and which corridors need urgent safety interventions. --- ## Features | Feature | Description | |---------|-------------| | KDE Hotspot Mapping | Kernel Density Estimation for accident concentration | | DBSCAN Clustering | Spatial clusters of high-frequency accident zones | | Severity Prediction | Random Forest model, Minor / Serious / Fatal | | Risk Factor Analysis | Hour-of-day, weather, road-type breakdown | | Interactive Heatmap | Folium map with annotated hotspot clusters | --- ## Tech Stack | Layer | Technology | |-------|-----------| | Geospatial | GeoPandas, Folium, Shapely | | Machine Learning | scikit-learn (Random Forest, KDE) | | Data | pandas, NumPy | | Visualisation | Matplotlib, Seaborn, Plotly | --- ## Project Structure ``` road-accident-hotspot/ ├── src/ │ ├── data_loader.py # Accident record ingestion & preprocessing │ ├── analysis.py # KDE hotspot detection, clustering, severity model │ └── visualize.py # Hotspot maps and risk factor charts ├── data/raw/ # FRSC accident logs, road network, weather data ├── models/ # Saved severity classifier ├── config.yaml # Parameters: bandwidth, thresholds, model config ├── main.py # Pipeline entry point └── requirements.txt ``` --- ## Quick Start ```bash git clone github.com cd road-accident-hotspot pip install -r requirements.txt python main.py ``` --- ## Data Sources - FRSC accident reports (GPS coordinates, …