Shrinkage and tree-based regression methods for the prediction of the live weight of Akkaraman sheep using morphological traits.

The prediction of live weight (LW) is of critical importance to farmers in a range of applications, including breeding and monitoring animal growth. In this context, Ridge, Least Absolute Shrinkage and Selection Operator (LASSO), and Elastic Net as shrinkage methods, and Classification and Regression Trees (CART) and Random Forest (RF) as tree-based regression methods were used in this study to predict LW of Akkaraman Sheep at 6-month age using sex, birth weight (BW) and some morphological traits such as withers height (WH), chest depth (CD), body length (BL), chest width (CW), rump height (RH), and chest circumference (CC). The dataset of 100 sheep, consisting of 44 males and 56 females, was randomly divided into training and test sets with a ratio of 80% and 20%, respectively. 10-fold cross-validation method was implemented using the training set to obtain optimum regression models and avoid overfitting. A test set was used to compare the prediction performance of regression methods based on various comparison criteria. Results revealed that LW was significantly correlated with all morphological traits and BW with coefficients ranging from 0.216 to 0.757. RF outperformed the other regression models with a coefficient of determination value (R²) of 0.865, followed by Ridge (R² = 0.761), LASSO (R² = 0.755), Elastic Net (R² = 0.750), and CART (R² = 0.654). The results indicated that WH and CD contributed the most, while sex and BW contributed the least in constructing the optimum RF model. In conclusion, the use of RF is recommended for predicting the LW of Akkaraman sheep. These results can provide a data-driven approach to improve decision-making in animal breeding.