Application of time series to estimate the number of dengue cases in Cascavel-PR

Abstract

Dengue fever is one of the most important diseases today, threatening about half of the world's population, especially in tropical regions with tropical and subtropical climates, in urban and suburban areas. The number of disease cases and deaths has increased significantly, leading to greater social and economic impacts. A good strategy to control the vector and properly treat those infected is the adequate preparation of the relevant agencies. Therefore, strategies that can predict the occurrence of dengue outbreaks play a key role. In this study, several strategies were evaluated to estimate the number of positive dengue cases for the municipality of Cascavel, Paraná. The methods evaluated were moving average, exponential smoothing, ARIMA, SVM, MLP, random forests, and recurrent neural networks. The information used in the study was the number of positive cases from 2007 to 2014, mounth of occurance, average temperature, rainfall index, and relative humidity. The MLP achieved the best performance with a value of 3.163 for the root mean square error (RMSE). The Random Forest and moving average algorithms also performed the best, with an RMSE of 3.264 and 4.123, respectively.