codoom1/TB_Ghana_Research

TB forecasting pipeline for Ghana using IHME GBD 2023 data, comparing classical and machine-learning models to project incidence, mortality, and mortality-to-incidence ratio to 2030. # Ghana Tuberculosis Trend and Forecasting This project uses IHME Global Burden of Disease 2023 data for tuberculosis in Ghana from 1990 to 2023. The workflow prepares the raw GBD extract, performs exploratory trend and stationarity analysis, evaluates eight competing forecasting models, and forecasts the primary TB incidence, mortality, and mortality-to-incidence ratio endpoints to 2030. ## Repository Structure ```text . ├── data/ │ ├── TB_df/ # Raw IHME GBD export and citation │ └── processed/ # Clean long/wide modeling datasets ├── docs/ │ ├── methods_model_documentation.md # Methods, equations, rationale, citations │ └── methodology_manuscript.md # Manuscript-style methods/results summary ├── outputs/ │ ├── figures/ │ │ ├── trend/ # Initial trend-analysis figures │ │ ├── eda/ # EDA, stratified, ACF/PACF figures │ │ ├── evaluation/ # Holdout prediction and error figures │ │ ├── forecasts/ # Point forecast comparison figures │ │ └── uncertainty/ # Forecast uncertainty figures │ ├── forecasts/ # Final forecast CSVs │ ├── model_outputs/ # Test-period model predictions │ └── tables/ # Summary, diagnostics, metrics, forecasts ├── scripts/ │ ├── 00_eda_stationarity.py # EDA plots and ADF/KPSS stationarity tests │ ├── 01_prepare_and_explore.py # Data preparation and initial trend outputs │ ├── 02_forecast_classical.py # Classical-only reference workflow │ ├── 03_train_evaluate_all_models.py# Full 8-model training/evaluation workflow │ ├── 04_forecast_best_models.py # Final best-model and all-model forecasts │ ├── 05_forecast_uncertainty.py # Simulation-based forecast uncertainty │ ├── compare_lstm_architectures.py # LSTM architecture and differencing diagnostics │ ├── compare_tcn_transforms.py …