Malaria Angola CLI

Hybrid AI framework for operational malaria incidence forecasting across all 18 provinces of Angola.

Integrates multi-engine predictive models with real-time satellite climate data and epidemiological memory features to generate province-level forecasts, severity assessments, and intervention alerts.

Installation

npm install -g malaria-angola-cli

Usage

malaria predict Uige                       # Province forecast
malaria predict --all --export csv          # National forecast with export
malaria alerts --severity critical          # Active epidemiological alerts
malaria climate --all                       # Real-time climate monitoring
malaria train                               # Model training pipeline
malaria report                              # Model performance metrics

Overview

The framework employs a weighted ensemble of specialized prediction engines operating across different methodological paradigms. Each engine captures distinct aspects of malaria transmission dynamics — from seasonal decomposition and temporal autocorrelation structures to non-linear feature interactions identified through gradient-boosted tree ensembles.

Forecasts incorporate 75 engineered features spanning climate reanalysis data (NASA POWER), 484 WHO/GHO health system indicators, and epidemiological memory variables with configurable lag structures. The system automatically evaluates alert thresholds derived from entomological transmission models and generates province-specific intervention recommendations.

When connected to the trained model API, the ML engine achieves validated performance with walk-forward cross-validation across the 2012–2023 evaluation period. The ensemble architecture provides graceful degradation — local analytical engines maintain operational capability independent of API connectivity.

Research

Full methodology, validation framework, and findings:

Operational Malaria Forecasting in Angola Using Ensemble Models, Regional Clusters, and Epidemiological Memory Features

Author

Joaquim Carlos Timoteo

LinkedIn · GitHub

License

MIT