Predicting Malaria Epidemics in Burkina Faso With Gaussian Processes

Here we present a combined early warning system and malaria predictor that can predict the 13 week trajectory of malaria cases in an primary health facility in Burkina Faso. Using the extraordinarily high fidelity data taken from the Integrated e-Diagnostic Approach (IeDA) system that has been rolled out throughout Burkina Faso, we train a combination of Gaussian Processes and Random Forest Regressors to estimate the trajectory of malaria cases over a 13 week period. We calibrate and test our algorithm such that it can return robust 1 and 2σ one-tailed and two-tailed confidence bounds. Given our lowest threshold for an epidemic alert our algorithm has 30% precision with >  99% recall. This rises to >  99% precision and 5% recall for the high alert threshold. Our two-tailed predictions have an average 1σ and 2σ precision of 5 cases and 30 cases respectively. Funding Statement: This work was in part funded by Cloudera Foundation, the Marguerite Foundation and the Delta ITP institute, and technically supported by Cloudera Foundation and Tableau Foundation. Declaration of Interests: MISSING