lavin-01/maize-yield-capstone-project

Research-oriented web application predicting maize yield and market prices at Kenya's county level. Combines machine learning models, real-time data analytics, and an intuitive dashboard for agricultural producers and researchers. Includes county intelligence reports, prediction history, and model explainability features. # Kenya Maize Intelligence Research-oriented web application for predicting **maize yield** and **maize prices** at the Kenya county level. ## Quick start See **docs/LOCAL_DEVELOPMENT.md** for full instructions. ```powershell # 1. ML pipeline (clean data + train models) .\.venv\Scripts\python -m ml.run_pipeline # 2. Initialize database cd backend $env:PYTHONPATH="." ..\.venv\Scripts\python scripts\init_db.py # 3. Start API ..\.venv\Scripts\uvicorn app.main:app --reload --port 8000 # 4. Start React (new terminal) cd frontend npm run dev ``` Open localhost — login with `producer@demo.ke` / `producer123` ## Stack - **Frontend:** React + TypeScript + Tailwind + Recharts - **Backend:** FastAPI REST + JWT + SQLAlchemy - **Database:** SQLite (local dev) or PostgreSQL + pgAdmin 4 (Docker) - **ML:** scikit-learn, XGBoost — yield & price models with feature explainability ## Project layout ``` kenya-maize-intelligence/ ├── backend/ # FastAPI REST API ├── frontend/ # React dashboard ├── ml/ # Data cleaning & model training ├── data/ │ ├── raw/ # Your original datasets │ └── processed/ # Cleaned merged outputs └── docs/ # Documentation ``` ## Model results (honest evaluation, test ≥ 2023) | Task | Best model | R² | RMSE | |------|------------|-----|------| | Yield | Linear Regression | 0.28 | 0.86 t/ha | | Price | Linear Regression | 0.31 | 7.49 KES/kg | ## Data Raw data from your capstone folder. See docs/DATA_INVENTORY.md. ## Azure Deployment guide coming in `docs/LOCAL_DEVELOPMENT.md` when you're ready — we'll use your Azure Student account with App Service + PostgreSQL.