Momahmoses/land-tenure-conflict

ML risk assessment of land tenure conflicts across Nigeria's Middle Belt and North-Central states with a Streamlit dashboard. # Land Tenure Conflict Risk & Resolution Platform ML-powered risk assessment platform for land parcels across Nigeria's Middle Belt and North-Central states, identifying high-conflict zones and supporting mediators, governors, and land registries in prioritising interventions. --- ## Problem Statement Land conflicts kill thousands in Nigeria annually. The Middle Belt farmer-herder crisis and widespread land grabbing destabilise communities and deter investment. Without systematic risk scoring, mediators cannot prioritise scarce resources effectively. --- ## Features | Feature | Description | |---------|-------------| | 4-Tier Risk Classifier | Gradient Boosting, Low / Medium / High / Critical | | 5,000 Parcel Records | Covering 12 Middle Belt and North-Central states | | Conflict Type Breakdown | Farmer-herder, boundary, land grabbing, ethnic claim | | Documentation Score Analysis | Correlation between land registry quality and conflict risk | | Parcel-Level Risk Scoring | Probability outputs for individual parcels | | Interactive Dashboard | Streamlit app with spatial risk map | --- ## Tech Stack | Layer | Technology | |-------|-----------| | Machine Learning | scikit-learn (Gradient Boosting), pandas | | Geospatial | GeoPandas, Folium | | Dashboard | Streamlit, Plotly | | Data | NumPy, pandas | --- ## Project Structure ``` land-tenure-conflict/ ├── src/ │ ├── data_generator.py # Synthetic parcel dataset (5,000 records) │ ├── model.py # Gradient Boosting classifier, risk scoring │ └── visualize.py # Conflict map, risk distribution charts ├── streamlit_app.py # Dashboard entry point ├── .streamlit/config.toml ├── requirements.txt └── runtime.txt ``` --- ## Quick Start ```bash git clone github.com cd land-tenure-conflict pip install -r requirements.txt streamlit run streamlit_app.py ``` --- ## Data Sources - NLC (National Land Commission) parcel registry - State la …