Estimation of the SIR Model Parameters Using Neural Networks
Valter de Assis Moreno Jr - candidato ao título de mestre

In the last decades, dengue fever has become the most prevalent epidemic disease caused by an arborvirus. Its socio-economic impact has been especially overloading to developing countries, which struggle with the lack of appropriate resources and policies to contain the disease. Good planning has been essential to this end and dramatically benefits from outbreak forecasts. Over time, several deterministic and stochastic mathematical models of dengue epidemics have been proposed. However, the methods used to estimate their parameters usually require complex calculations and strong distributional assumptions that may not be realistic. The objective of this study was to develop a data-driven method to estimate the parameters of epidemiological models using Machine Learning and Artificial Neural Networks (ANNs) that could circumvent such demands. To accomplish this, we created a data set of infectives time series generated with SIR models using parameters derived from previous dengue epidemics and additional random noise. We used the data to train and validate several neural network configurations using the Mean Squared Error (MSE) and the Mean Absolute Percentage Error (MAPE) as the loss function. The test of the best models showed that the MAPE network tended to estimate SIR models that fitted the data better than the MSE network. We then applied the MAPE model to the time series of dengue epidemics that occurred in Brazilian state capitals between 2007 and 2020. The overall results indicate that ANN data-driven estimation methods can be used to fit a deterministic epidemiological model to noisy data, at least in cases where the dynamic processes that underlie the generation of observations are similar to those specified in the model.

Texto informado pelo autor.

Membros da banca
Flávio Codeço Coelho - FGV EMAp
Luiz Max Fagundes de Carvalho - FGV EMAp
Américo Barbosa da Cunha Junior - UERJ
Marcelo Ferreira da Costa Gomes - Fiocruz
High contrast

Esse site usa cookies

Nosso website coleta informações do seu dispositivo e da sua navegação e utiliza tecnologias como cookies para armazená-las e permitir funcionalidades como: melhorar o funcionamento técnico das páginas, mensurar a audiência do website e oferecer produtos e serviços relevantes por meio de anúncios personalizados. Para mais informações, acesse o nosso Aviso de Cookies e o nosso Aviso de Privacidade.