View article

Based solely on the dengue confirmed-cases of six densely populated urban areas in Brazil, distributed along the country, we propose in this paper regularized linear and nonlinear autoregressive models for one-week ahead prediction of the future behaviour of each time series. Though exhibiting distinct temporal behaviour, all the time series were properly predicted, with a consistently better performance of the nonlinear predictors, based on MLP neural networks. Additional local information associated with environmental conditions will possibly improve the performance of the predictors. However, without including such local environmental variables, such as temperature and rainfall, the performance was proven to be acceptable and the applicability of the methodology can then be directly extended to endemic areas around the world characterized by a poor monitoring of environmental conditions. For …