Application of Artificial Neural Network and the Monomolecular Model in Describing the Relationship Between Body Weight Gain and Metabolizable Energy Intake in Egg-Type Pullets.

BACKGROUND: Modelling growth allows nutritionists and poultry researchers to predict dynamic or daily nutrient needs more precisely than using fixed requirements. OBJECTIVES: This study evaluated the monomolecular model and artificial neural network (ANN) to describe the relationship between metabolizable energy (ME) intake and body weight gain (BWG) in commercial strains of different egg-type pullets. METHODS: A multi-layer feed-forward perceptron neural network structure was used to construct the ANN model. The best-fitted network on input data to predict BWG from ME intake in all egg-type pullets' strains was obtained with one neuron in the input layer, three neurons in the first hidden layer, two neurons in the second hidden layer and one neuron in the output layer, which was written as 1-3-2-1. RESULTS: The relationship between ME intake and BWG in different egg-type pullet strains was predicted with very high accuracy by both ANN (R2 adj = 99.47-99.99) and the monomolecular model (R2 adj = 98.48-99.72). The maintenance energy requirements (134.2-165.8 kcal/kg BW) and efficiencies of NE utilization for growth (2.23-4.00 kcal/g of BW) estimated by the monomolecular model are consistent with previously reported values for poultry. Meta-analysis of the parameters estimated by the monomolecular model (a, b and c) revealed significant differences among strains (p < 0.01), suggesting that these strain-specific growth responses may be linked to genetic diversity. CONCLUSIONS: This study demonstrated that both the monomolecular model and ANN approaches effectively described the relationship between ME intake and BWG in egg-type pullets.