Estudo preliminar para predição de consumo de nutrientes e desempenho de ovinos deslanados em confinamento baseado em aprendizagem de máquina

• Ano de defesa: 2023
• Autor(a) principal: COSTA, Hadja Lorena Rangel Uchôa Cavalcanti de Menezes
• Orientador(a): VÉRAS, Antonia Sherlânea Chaves
• Banca de defesa: CAMPELO, Cláudio Elisio Calazans, GOMES, Renata Valéria Regis de Sousa
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal Rural de Pernambuco
• Programa de Pós-Graduação: Programa de Pós-Graduação em Zootecnia
• Departamento: Departamento de Zootecnia
• País: Brasil
• Palavras-chave em Português: Desempenho animal; Ovino; Aprendizagem de máquina; Inteligência artificial; Consumo de nutriente
• Área do conhecimento CNPq: CIENCIAS AGRARIAS::ZOOTECNIA
• Link de acesso: http://www.tede2.ufrpe.br:8080/tede2/handle/tede2/9169

Once the main reason for confinement of livestock animals is to produce quality meat to supply the domestic market with regular supply and at the lowest production cost, the use of an Artificial Neural Network (ANN) system that allows the prediction of animal performance based on initial inputs and historical data seems promising, enabling low-performing animals to be identified in advance and discarded to maximize production. Given the importance of this information, as well as nutrient consumption by animals, the present study aims to investigate the use of ANN as predictors of nutrient intake and, from the chemical composition of the diet, predict the performance of shorn sheep in confinement. The database was obtained from information from scientific experiments conducted in the sheep sector of the Department of Animal Science at the Federal Rural University of Pernambuco. To test the efficacy of ANN in predicting desired characteristics, nine other prediction techniques with distinct features were explored to evaluate different scenarios and observe the best technique for each situation. Dealing with all predicted points, such as average daily intake, average daily gain, and final body weight, the ANN demonstrated a performance well below expectations when compared to other techniques included in the analysis. This may be a consequence of the low volume of data available for training, verification, and checking of the prediction models. Further studies are needed to explore other promising prediction tools for animal science as a whole. In summary, although ANNs may be a promising technique in the field of animal science, it is important to carefully evaluate the selection of variables and appropriate techniques for each specific case to obtain more accurate and reliable results.