Machine Learning Techniques Using Enviromental Data from Remote Sensing Applied to Modeling Dengue Risk in Brazil

Mosquitoes are the most important vectors of human diseases in tropical countries. In particular, Aedes aegypti is the main vector for Chikungunya, Dengue and Zika in South America and Asia. In recent years, transmission of these diseases has increased predominantly in urban areas and has become a major public health problem. This study presents an algorithm for dengue risk prediction in Brazil. It is based in machine learning techniques and uses environmental and socioeconomic variables to predict the risk of dengue propagation all over Brazil, at municipality level. Data from 2010 to 2013 of monthly dengue episodes distribution have been obtained from the Notifiable Diseases Information System (SINAN), developed by the Brazil ministry of health. In addition, environmental data had been obtained from the distributed products of earth observation satellite missions that acquire globally and periodically data. Three types of machine learning algorithms have been applied and compared. The best results have been obtained when using random forests.