Complementary Framework

Disease scope

We currently support two diseases

Malaria and Tuberculosis, with time-series forecasting and model testing APIs.

Malaria

Monthly surveillance with district granularity. Models focus on confirmed cases by age and sex using lagged features, with optional environmental exogenous variables.

Typical inputs

  • History: date, district, target counts
  • Optional exogenous: rainfall, temperature, NDVI
  • Granularity: monthly (MS)

Targets

  • Confirmed cases by age/sex cohorts
  • Facility or district totals

Available models

Detected automatically from the API.

Malaria surveillance visualization

Quick API examples

# Predict (tabular)
                                    curl -s -X POST ${API_BASE}predict/<malaria_model> \
                                    -H "Content-Type: application/json" \
                                    -d '{"input": [[f1,f2,...,f12]]}'

                                    # Forecast (monthly, panel by district)
                                    curl -s -X POST ${API_BASE}forecast/<malaria_model> \
                                    -H "Content-Type: application/json" \
                                    -d '{
                                            "h": 3,
                                            "history": [
                                            {"date":"2024-10-01","district":"Abim District","y":0.0},
                                            {"date":"2024-11-01","district":"Abim District","y":0.0},
                                            {"date":"2024-12-01","district":"Abim District","y":1.0}
                                            ]
                                        }'

Model types

Random Forest and XGBoost. Forecasting uses lag features and optional exogenous signals.

API endpoints

/predict/<model>, /forecast/<model>, /evaluate/<model> (if implemented by the model).

Data governance

Write actions require an API key. Only aggregated, de-identified data should be sent to the API.