Spatiotemporal forecasting for dengue, chikungunya fever and Zika using machine learning and artificial expert committees based on meta-heuristics

Abstract

PurposeDengue is considered one of the biggest public health problems in recent decades. Climate and demographic changes, the disorderly growth of cities and international trade have brought new arboviruses such as chikungunya and Zika. Control of arboviruses depends on control of the vector: the Aedes aegypti mosquito.ObjectiveIn this work, we propose a methodology for building disease predictors capable of predicting infected cases and locations based on machine learning. We also propose an artificial experts committee based on meta-heuristic methods to detect the most relevant risk factors. MethodAs a case study, we applied the methodology to forecast dengue, chikungunya and Zika, with data from the City of Recife, Brazil, from 2013 to 2016. We used arboviruses cases data and climatic and environmental information: wind speeds, temperatures and precipitation. ResultsThe best prediction results were obtained with 10-tree Random Forest regression, with Pearson’s correlation above 0.99 and RMSE (%) below 6%. Additionally, the artificial experts committee was able to present the most relevant factors for predicting cases in each two-month period.ConclusionThe spatiotemporal prediction results showed the evolution of arboviruses, pointing out as major focuses on both regions richer in urban green areas and low-income neighborhood with irregular water supply. Determining the most relevant factors for prediction, as well as the spatial distribution of cases, can be useful for the planning and execution of public policies aimed at improving the health infrastructure and planning and controlling the vector.