Comparison of logistic and neural network models to fit to the egg production curve of White Leghorn hens

Abstract

Neural networks are capable of modeling any complex function and can be used in the poultry and animal production areas. The aim of this study was to investigate the possibility of using neural networks on an egg production data set and fitting models to the egg production curve by applying 2 approaches, one using a nonlinear logistic model and the other using 2 artificial neural network models [multilayer perceptron (MLP) and radial basis function]. Two data sets from 2 generations of a White Leghorn strain that had been selected mainly for egg production were used. In the first data set, the mean weekly egg-laying rate was ascertained over a 54-wk egg production period. This data set was used to adjust and test the logistic model and to train and test the neural networks. The second data set, covering 52 wk of egg production, was used to validate the models. The mean absolute deviation, mean square error, and R(2) were used to evaluate the fit of the models. The MLP neural network had the best fit in the test and validation phases. The advantage of using neural networks is that they can be fitted to any kind of data set and do not require model assumptions such as those required in the nonlinear methodology. The results confirm that MLP neural networks can be used as an alternative tool to fit to egg production. The benefits of the MLP are the great flexibility and their lack of a priori assumptions when estimating a noisy nonlinear model.