Redes neuronales artificiales para la predicción de la masa corporal de pollos

• Autor(a) principal: Ferraz, Patrícia Ferreira Ponciano
• Data de Publicação: 2019
• Outros Autores: Yanagi Junior, Tadayuki, Julio, Yamid Fabián Hernández, Ferraz, Gabriel Araújo e Silva, Cecchin, Daiane
• Tipo de documento: Artigo
• Idioma: spa
• Título da fonte: Repositório Institucional da UFLA
• Texto Completo: http://repositorio.ufla.br/jspui/handle/1/40313
• topic: Animal welfare; Artificial intelligence; Broiler - Thermal comfort; Bienestar animal; Inteligencia artificial; Pollos - Confort térmico; Bem-estar animal; Frango - Conforto térmico

The thermal environment inside a broiler house has a great influence on animal welfare and productivity during the production phase. Thus, the aim of this study was to predict body mass of chicks from 2 to 21 days of age when subjected to different intensities (27, 30, 33 and 36°C) and duration (1, 2, 3 and 4 days starting on the second day of life) using artificial neural networks (ANN). This experiment was conducted at Lavras, MG, Brazil. It was used 210 chicks of both sexes, from 1st to 22nd days of life. The chicks were raised inside four climate-controlled wind tunnels. Daily the weight of all the chicks was measured to know the daily body masses. The input variables were dry-bulb air temperature, duration of thermal stress, chick age, and the output variable was the daily body mass of chicks. A database containing 840 records was used to train (70% of data), validate (15%) and test (15%) of models based on artificial neural networks (ANN). Between these models, the ANN was accurate in predicting the BM of chicks from 2 to 21 days of age after they were subjected to the input variables, and it had an R² of 0.9992 and a standard error of 5,23 g. This model enables the simulation of different scenarios that can assist in managerial decision-making, and it can be embedded in the heating controls.