Ghentlesteve/malaria-risk-nigeria

# Nigeria State-Level Malaria Risk Stratification ## Overview This project builds a machine learning pipeline to classify malaria risk at the state level across Nigeria, using DHS/MIS survey indicators combined with climate variables (rainfall and vegetation index). The goal is to produce an operational risk stratification tool that supports national malaria program planning and resource allocation. A secondary module flags states with low reporting reliability, providing a data-confidence layer on top of the risk predictions — drawing on established data quality assessment (DQA) frameworks used in DHIS2-based health information systems. **Live demo:** _[link added after deployment]_ --- ## Problem Statement Nigeria bears one of the world's highest malaria burdens. National malaria programs must prioritize limited resources — bed nets, indoor spraying, treatment stockpiles — across 36 states and the FCT. Doing this well requires knowing not just current prevalence, but which states are likely to remain high-risk or transition into high-risk zones, and whether the data supporting those estimates is reliable. This project addresses both questions: 1. **What is the predicted malaria risk class for each state?** (Low / Medium / High) 2. **How confident should we be in that prediction, given reporting data quality?** --- ## Data Sources | Source | Description | Granularity | |--------|-------------|-------------| | DHS Program / HDX | Nigeria subnational malaria indicators (ITN coverage, RDT/microscopy prevalence, ANC indicators) | State × Survey year (2003–2024) | | CHIRPS | Monthly rainfall estimates | State-level spatial aggregation | | MODIS NDVI | Vegetation index (proxy for mosquito habitat) | State-level spatial aggregation | --- ## Methodology 1. **Data ingestion & cleaning** — pivot DHS indicators from long to wide format, handle missing survey years, merge climate variables 2. **Feature engineering** — ITN coverage rates, ANC intervention indicato …