tina-tynash/cholera-zim-forecast

Cholera forecasting for Zimbabwe - ML ensemble + security # 🦠 Cholera Forecasting in Zimbabwe > **Enhancing Cholera Forecasting in Zimbabwe through Interdisciplinary Data-Driven Approaches and Accessibility of Epidemiological Data** An open-source, reproducible ML system for cholera outbreak prediction in Zimbabwe. Integrates epidemiological surveillance, climate data, and socioeconomic indicators into an ensemble forecasting pipeline with an accessible interactive dashboard. ## Quick Start ```bash git clone github.com cd cholera-zim-forecast pip install -r requirements.txt python data/synthetic/generate_synthetic.py python src/data/etl.py streamlit run src/app/streamlit_app.py ``` ## Docker ```bash docker-compose -f docker/docker-compose.yml up --build # Dashboard: localhost # API: localhost ``` ## Key Results | Model | MAPE (%) | RMSE | Notes | |---|---|---|---| | ARIMA baseline | 31.4 | 124.2 | No climate features | | Prophet | 19.8 | 87.4 | + rainfall regressor | | XGBoost | 16.3 | 71.8 | + SHAP interpretability | | LSTM | 17.1 | 76.3 | 12-week sliding window | | **Ensemble (Ours)** | **13.7** | **58.9** | **18% vs best single model** | ## Architecture ``` Raw Data (HDX, ERA5, ZimStat) │ ▼ ETL Pipeline ──► Feature Engineering ──► SQLite/PostgreSQL │ ├──► Prophet ──┐ ├──► XGBoost ──┼──► Stacked Ensemble ──► Forecasts └──► LSTM ─────┘ │ ▼ Streamlit Dashboard + FastAPI ``` ## Citation ```bibtex @article{cholera_zim_2026, title = {Enhancing Cholera Forecasting in Zimbabwe}, author = {Your Name}, year = {2026} } ``` MIT License | See docs/ethics_statement.md for data ethics