Use este identificador para citar ou linkar para este item: http://repositorio.ufla.br/jspui/handle/1/30474
Título: Estudo comparativo de técnicas de inteligência computacional para estimação dos requerimentos energéticos de gado bovino
Título(s) alternativo(s): Comparative study of computational intelligence techniques for estimation of energy requirements of cattle
Autores: Ferreira, Danton Diego
Gionbelli, Mateus Pies
Ferreira, Danton Diego
Cerqueira, Augusto Santiago
Lacerda, Wilian Soares
San Vito, Elias
Palavras-chave: Inteligência computacional
Redes Neurais
Lógica fuzzy
Bovino - Exigências energéticas
Bovino - Alimentação e rações
Consumo de Energia Metabolizável (CEM)
Computational intelligence
Neural networks
Fuzzy logic
Cattle - Energy Requirements
Cattle - Feeding and feeds
Metabolizable energy
Data do documento: 24-Set-2018
Editor: Universidade Federal de Lavras
Citação: LIMA, R. R. de. Estudo comparativo de técnicas de inteligência computacional para estimação dos requerimentos energéticos de gado bovino. 2018. 93 p. Dissertação (Mestrado em Engenharia de Sistemas e Automação)–Universidade Federal de Lavras, Lavras, 2018.
Resumo: Brazil is the second largest exporter of beef in the world and the increase in consumption and exports, will demand an increase in production. Due to environmental issues, it is important that this increase occurs without significantly increasing the production area. The key to achieve this goal lies in an accurate knowledge of the nutritional requirements and the food composition offered to the cattle. Techniques based on Computational Intelligence (CI) are spreading in animal sciences and shown as effective tools for the most diverse applications. In this Master’s Dissertation, were developed models capable of estimating the energy requirements of cattle, one of the bases for the elaboration of efficient diets for the animals. Techniques based on Artificial Neural Network (ANN) and Fuzzy Inference Systems (FIS), foundations of CI, were developed in this work as alternatives to classical methods. The techniques covered were: Multilayer Perceptron (MLP) Networks with Backpropagation training based on Squared Conjugate Gradient (SCG), Generalized Regression Neural Networks (GRNN) and Adaptive Neuro-Fuzzy Inference Systems (ANFIS). All techniques were used to estimate the Metabolizable Energy Intake (MEI) from a database of 840 animals from 31 studies. The results were analyzed statistically and compared with each other, with the classical modeling method, based on Multiple Regressions (MR) and with the equations suggested by BR-CORTE System. The parameters used for the modeling were: gender, genetic group, feeding system, Average Empty Body Weight (AEBW) and Empty Body Weight Gain (EBWG). The proposed models were able to overcome the average performance of classical technique in five of the eight metrics used to statistically evaluate the resulting models, presenting a significant gain by the Tukey Test in three. The best methodology was the MLP, with superior results in relation to linear correlation, determination index and concordance correlation coefficient. The results confirm the validity of CI methodologies for the prediction of energy requirements in beef cattle.
URI: http://repositorio.ufla.br/jspui/handle/1/30474
Aparece nas coleções:Engenharia de Sistemas e automação (Dissertações)



Os itens no repositório estão protegidos por copyright, com todos os direitos reservados, salvo quando é indicado o contrário.