EPIC Health Systems home
Reach Us
Digital health, AI & innovation

A machine learning forecast for dengue outbreaks

Dengue outbreaks in Ethiopia have become more frequent, and the health system has no tool to predict when and where they will occur. This project builds and tests a model that forecasts short-term dengue risk at district level from surveillance and climate data.

CountryEthiopia
PeriodFebruary 2026 – January 2027
StatusOngoing
Health topicDengue
01

Challenge

Since the first major outbreak in Dire Dawa in 2013, dengue has spread across eastern and northeastern Ethiopia. Response usually starts after cases rise, which limits what mosquito control and public messaging can achieve.

02

Approach

District case counts for 2013 to 2023 are combined with weekly rainfall, temperature and humidity, and with population and land cover data. Random forest, support vector regression and gradient boosting models are trained with walk-forward validation and compared on how well they predict weekly incidence and outbreak weeks. The results will be reviewed with public health officers.

03

Expected outputs

A validated forecast model, evidence on which climate and surveillance variables best predict outbreaks, and guidance on how forecasts could feed routine surveillance.

Work with us.

We take enquiries from governments, development partners, research institutions and implementers on research, advisory and training engagements.

Get in touch
Partners & collaborators

Who we work with