Dados do Trabalho

Título

PREDICTING THE OCCURRENCE OF TEGUMENTARY LEISHMANIASIS IN MARANHÃO, BRAZIL: A TIME-SERIES ANALYSIS USING A HYBRID MODEL

Introdução

The occurrence of tegumentary leishmaniasis (TL) has increased in recent decades, becoming an important public health problem. Its early detection through predictive methods enables the development of better intervention and control strategies and better allocation of resources.

Objetivo(s)

The objective was to develop a hybrid predictive method between a seasonal autoregressive linear integrated moving average (SARIMA) model and nonlinear autoregressive neural network with exogenous input (NARNNX) to estimate the occurrence of LC in Maranhão, Brazil.

Material e Métodos

Monthly data of cases of TL notified and confirmed in the Notifiable Diseases Information System in the period from January 2007 to December 2019 were used. Thus, one has 156 months of observations over a 13-year period for the analysis. Non-hybrid SARIMA and hybrid SARIMA-GRNN (generalized regression neural network), SARIMA-NARNN (non-linear autoregressive neural network) and SARIMA-NARNNX (non-linear autoregressive neural network with exogenous input) approaches were employed for performance comparison. The lowest mean absolute percentage error (MAPE), mean error rate (MER), mean absolute error (MAE), and root mean square error (RMSE)
were used to compare the predictive performances of the proposed models. After the best model was chosen, forecasting was performed for the period from January 2020 to December 2023.

Resultados e Conclusão

In this observational study, 25001 cases of the disease were reported, with a seasonal peak in October and November each year. The combined SARIMA-NARNNX (5,1,4)12 approach showed the best fit. In 2023, TL will show a decreasing trend, but still maintaining an important record of cases. The hybrid SARIMA-NARNNX model proved adequate to simulate and predict the trend of TL in Maranhão,
Brazil.

Palavras-chave

Tropical diseases, Epidemiology, Time Series.

Área

Eixo 06 | Protozooses