|Dozier Iii, William
Submitted to: Poultry Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/15/2005
Publication Date: 4/15/2006
Citation: Roush, W.B., Dozier III, W.A., Branton, S.L. 2006. Comparison of gompertz and neural network models of broiler chickens. Poultry Science. 85:794-797.
Interpretive Summary: The accurate and precise fitting of observed growth data with growth models is necessary to gain a mathematical understanding of the biological relationship between animals and their physical environment. Animal models are commonly developed using regression analysis. Artificial neural networks offer an alternative to regression analysis in that neural network models have proven to be superior in prediction of responses. In this study the daily responses of broilers were modeled by regression using the Gompertz equation. The Gompertz equation is a commonly accepted nonlinear regression growth model. Neural networks are computerized versions of connected neurons used to mimic the way that the brain learns. The body weight data were divided into two sets - one to develop the model and a second to validate the model. The appropriateness of the Gompertz and neural network models were evaluated using error measurements commonly used in evaluating forecasting models. The Gompertz model was found to underestimate the observed responses while the neural network models produced little or no overestimation of the body weight responses.
Technical Abstract: Neural networks offer an alternative to regression analysis for biological growth modeling. There is very little research that has been conducted to model animal growth using artificial neural networks. Twenty-five male chicks (Ross x Ross 308) were raised in an environmental chamber. BW were determined on a daily basis. Feed and water were provided ad libitum. The birds were fed a starter diet (23% CP 3,200 kcal ME/kg) from 0 to 21 d, and a grower diet (20% CP and 3,200 kcal ME/kg) from 22 to 70 d. Dead and female birds were not included in the study. Average BW of 18 birds were used as the data points for the growth curve to be modeled. Training data consisted of every other d weights starting with the first d. Validation data consisted of BW at all other age periods. Comparison was made between the modeling by the Gompertz nonlinear regression equation and neural network modeling. Neural network models were developed with the Neuroshell Predictor. Accuracy of the models were determined by Mean Square Error (MSE), Mean Absolute Deviation (MAD), Mean Absolute Percentage Error (MAPE), and Bias. The Gompertz equation was fit for the data. Forecasting error measurements were based on the difference between the model and the observed values. For the training data, the lowest MSE, MAD, MAPE and bias were noted for the neural developed neural network. For the validation data, the lowest MSE, MAD, and MAPE were noted with the genetic algorithm developed neural network. Lowest bias was for the neural developed network. As measured by bias, the Gompertz equation under estimated the values while the neural and genetic developed neural networks produced little or no overestimation of the observed BW responses. Past studies have attempted to interpret the biological significance of the estimates of the parameters of an equation. However, it may be more practical to ignore the relevance of parameter estimates and focus on the ability to predict responses.