Título: Deep Learning-Based Ovitrap Spatial Dynamics Analysis for Arbovirus Vector Monitoring.
Autores: Ignacio Sanchez-Gendriz, Matheus Diniz, A. D. Doria Neto, Rodrigo Moreira Pedreira, Ion de Andrade, R. A. de Medeiros Valentim
Resumo: Dengue is a significant global health issue, affecting millions of people annually and imposing substantial socioeconomic burdens. Effective disease control relies on monitoring the population of Aedes aegypti mosquitoes, the primary vector of dengue. One surveillance method involves counting the eggs laid by these mosquitoes in spatially distributed ovitraps. This study focuses on the application of computational methods to forecast dengue vector populations. We analyze a four-year (2016- 2019) database from 397 ovitraps distributed across Natal, RNBrazil, with a weekly sampling frequency. Our objective is to develop accurate machine learning (ML) models that can predict the egg density index (EDI) at a fine-grained spatial resolution, aligned with zoonosis interventions. To preprocess the dataset obtained from the ovitraps, we employ spatial smoothing techniques and aggregation. The preprocessed data is then used to train ML models, including recurrent deep learning (DL) models, enabling accurate forecasting of the EDI. This approach shows promise for monitoring and preventing arbovirus outbreaks. Our findings demonstrate the effectiveness of spatial smoothing and aggregation as preprocessing steps for reducing randomness and noise in the dataset. The recurrent DL models exhibit high forecasting accuracy, thereby validating their utility in arbovirus monitoring and prevention efforts.
Palavras-chave: Machine learning, deep learning, arboviruses forecasting, dengue, ovitraps.
Páginas: 7
Código DOI: 10.21528/CBIC2023-056
Artigo em pdf: CBIC_2023_paper056.pdf
Arquivo BibTeX: CBIC_2023_056.bib