Título: Rede neural artificial para predição de brucelose bovina a partir de dados desbalanceados
Autores: Caio Donizetti Queiroz Alves, Danton Diego Ferreira, Danielle Abreu Fortunato, Christiane Maria Barcellos Magalhaes da Rocha
Resumo: The expressiveness of Brazilian livestock is unquestionable. According to data from the United States Department of Agriculture (USDA), in 2021 Brazil was the worlds largest exporter of beef. Bovine brucellosis is one of the most worrying diseases for the sector. In Brazil, bovine brucellosis causes annual losses of around 448 million dollars. Several factors threaten the establishment of actions of the current animal defense programs in Brazil, the main ones being: lack of distinct guidelines for the diagnosis of brucellosis cases, infected animals remain asymptomatic when infected, extensive Brazilian territory and large number of herds. The use of Artificial Neural Networks (ANNs) can be very useful in health and epidemiological surveillance services, helping to screen properties with different risks for the disease. The objective of this work is the development of ANN with class balancing techniques and selection of variables via genetic algorithm, for the classification and segregation of bovine herds, regarding seroprevalence for brucellosis. Five ANNs were designed combining different approaches of class balancing technique and variable selection, in order to compare which approach would perform better results. The results showed that ANN combined with the variable selection technique, via Genetic Algorithms, and class balancing, is a promising approach.
Palavras-chave: Artificial Neural Networks, brucellosis, class balancing, feature selection
Páginas: 8
Código DOI: 10.21528/CBIC2023-108
Artigo em pdf: CBIC_2023_paper108.pdf
Arquivo BibTeX: CBIC_2023_108.bib