Dev-Romm/Drought-Predictor

A Prophet forecasting web app for drought prediction in Turkana County # Drought Predictor A web application that provides NDVI-based drought forecasting for Turkana County, Kenya. The system helps pastoralists make informed decisions about livestock management and migration planning by visualizing historical vegetation conditions and predicting future drought events. ## Overview The Drought Predictor consists of: - **React Frontend**: Interactive dashboard with data visualization, forecasts, and drought alerts - **FastAPI Backend**: REST API server with ML model inference (Prophet and LSTM) - **ML Models**: Pre-trained Prophet and LSTM models for NDVI forecasting ## Features - Historical NDVI data visualization (2017-2026) - Dual forecasting models (Prophet and LSTM) - Configurable forecast horizons (2, 4, or 6 weeks) - Drought severity classification (Normal, Alert, Alarm, Emergency) - Pastoralist-friendly insights and recommendations - Interactive charts with confidence intervals - Real-time drought event markers ## Technology Stack **Frontend:** - React 18+ - Recharts for data visualization - Axios for API requests **Backend:** - FastAPI - Prophet (fbprophet) for time series forecasting - TensorFlow/Keras for LSTM inference - Pandas for data processing - Mangum for AWS Lambda compatibility ## Prerequisites **Frontend:** - Node.js 16+ and npm **Backend:** - Python 3.10+ - pip ## Setup Instructions ### Backend Setup 1. Navigate to the server directory: ```bash cd server ``` 2. Create a virtual environment: ```bash python -m venv venv ``` 3. Activate the virtual environment: - Windows: `venv\Scripts\activate` - macOS/Linux: `source venv/bin/activate` 4. Install dependencies: ```bash pip install -r requirements.txt ``` 5. Create a `.env` file based on `.env.example`: ```bash cp .env.example .env ``` 6. Edit `.env` to configure your environment variables 7. Ensure model files are present in `server/models/`: - `ndvi_prophet_model.pkl` - `ndvi_lstm_model.keras` - `ndvi_lstm_scaler.pkl` 8. Ensure the CSV data file exi …