Comparison of some non-linear functions to describe the growth for Linda geese with CART and XGBoost algorithms

The aim of this study was to determine the best non-linear function describing growth Linda goose breed. To achieve aim, five functions, such as exponential, logistic, von Bertalanffy, Brody and Gompertz, were employed define live weight-age relationship for male female geese. In study, 2 397 body records from 75 females 66 males collected three days 17 weeks age evaluated using "easynls" "nlstools" packages modelling in R software. Each model analysed weight all geese separately females. measure predictive quality functions used individually here, goodness fit criteria, coefficient determination (R2), adjusted (R2adj), root mean square error (RMSE), Akaike's information criterion (AIC) Bayesian (BIC) implemented. Among Bertalanffy gave curve Based on "rpart", "rpart.plot", "caret" packages, CART XGBoost algorithms specified prediction at parameters sex factor. produced better results superiority compared with algorithm. conclusion, it could be suggested that might help breeders appropriate slaughtering time, feeding regimes, overcome flock management problems. algorithm present a good reference establish breed standards selection strategies breeding purposes.