Isaac25-lgtm/uganda-malaria-early-warning

District-level malaria surge forecasting and early warning for Uganda's 146 districts, using DHIS2 HMIS 033b surveillance, CHIRPS rainfall and ERA5-Land climate data. MSc research proof-of-concept. Uganda Malaria Early Warning System District-Level Malaria Surge Prediction and Early Warning Using Climate Data in Uganda An MSc research proof-of-concept integrating routine HMIS surveillance, satellite-derived climate predictors and validated machine-learning models into an auditable early-warning dashboard. --- ## Overview This system turns two routinely available data streams — weekly DHIS2 malaria surveillance and satellite-derived climate data — into forward-looking district risk information for all **146 districts and cities in Uganda**. For every district it produces: - **Case forecasts** at 4, 8 and 12 weeks ahead, from a Random Forest model validated on six years of national data; - **A surge warning** against a district- and season-specific epidemic threshold, so a "high" week in Kampala is judged by Kampala's own seasonal baseline; - **A risk classification** rendered on a national choropleth map and a searchable district table; - **The evidence behind each number** — validation metrics, feature importance, data-quality coverage and threshold logic are all exposed in the interface rather than hidden in the model. At the four-week horizon the model detects **73% of surge weeks** and cuts forecast error by **30% relative to the naive benchmark** that current practice most closely resembles. Uganda's districts presently build endemic channels manually in Excel, with no climate input and no forecast; this system automates that baseline and extends it three months ahead. The analytical workflow is auditable in source and protected by automated tests and a deterministic SHA-256 integrity fingerprint over the published forecasts. Complete result reproduction additionally requires the governed surveillance data and production artifacts, which cannot be redistributed in this public repository. --- ## Table of contents - 1. Research context - 2. Contribution - 3. Data sources - 4. Methodology - 5. Results - 6. System architecture - 7. App …