Growth modeling of Lori sheep using artificial neural networks and nonlinear regression.

This study evaluates growth performance in Lori sheep using two modeling approaches: traditional nonlinear regression and artificial neural networks (ANNs). Body weight data were collected from 210 lambs at monthly intervals from birth to six months of age. Six nonlinear growth models including Brody, von Bertalanffy, Gompertz, Logistic, Verhulst, and negative exponential were compared with two neural network architectures: multilayer perceptron (MLP) and radial basis function (RBF). Model performance was assessed using the coefficient of determination (R²), root mean square error (RMSE), Akaike information criterion (AIC), and correlation between observed and predicted values. Among the nonlinear models, the Brody function provided the best fit, accurately capturing sigmoidal growth behavior and offering biologically interpretable parameters such as asymptotic mature weight (A) and maturity rate (k). Males exhibited higher growth rates and asymptotic weights than females, indicating sexual dimorphism. Growth velocity peaked early and declined with age, consistent with physiological development patterns. Artificial neural networks achieved high predictive accuracy, with correlation coefficients exceeding 0.996. Although these models effectively captured nonlinear growth dynamics, traditional models remain preferable in farm settings due to their interpretability and ease of use. The Brody model also proved reliable for predicting live weight at later stages using early body weight data, supporting improved decisions in nutrition planning, health monitoring, and breeding selection. For applications focused on growth curve monitoring, neural networks offer a practical alternative capable of automating evaluations and delivering accurate insights without requiring statistical expertise. Integrating biologically grounded models with machine learning techniques may enhance predictive accuracy and livestock management. Future research should explore hybrid modeling frameworks and incorporate environmental and nutritional variables to further optimize growth prediction.