Use Of Predictive Recurring Neural Network Models In The Prediction Of Dengue Cases (Previsão de casos de dengue utilizando modelos preditivos derede neural recorrente - RNN)

Resumo

https://doi.org/10.5281/zenodo.15699989

This research addresses the prediction of dengue outbreaks using statistical data, aiming to develop a predictive model that assists in the early identification of dengue outbreaks in the State of São Paulo. A quantitative approach was employed, utilizing statistical analysis techniques and mathematical models to process and interpret data collected from epidemiological and meteorological sources. Factors such as temperature, humidity, age, population, gender, waste collection, and access to treated water were analyzed, as well as the incidence of dengue cases in different regions of the State between the years 2010 and 2024. The methodology included the application of time series in predictive models developed using Recurrent Neural Networks (RNN) with three types of layers: Simple RNN, Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU). The results indicated that the combination of climatic variables and historical dengue data allows for the accurate prediction of new outbreaks. It was concluded that the application of statistical models in predicting dengue outbreaks can be a valuable tool for public health authorities, enabling the implementation of preventive and control measures more effectively and promptly, thus contributing to reducing the impact of dengue in affected communities.