Body Weight Prediction in Karayaka Lambs Using Morphometric Measurements: A Comparison of Regression and Machine Learning Approaches

Body weight is one of the important phenotypic traits in sheep breeding for evaluating growth performance, planning flock management practices, and determining economic efficiency. In this study, biometric and machine learning approaches were jointly employed to predict body weight in Karayaka lambs using morphometric characteristics. The research material consisted of a total of 150 Karayaka lambs, including 75 males and 75 females, raised in a private enterprise located in the Erbaa district of Tokat province. The study evaluated body weight (BW), heart girth (HG), abdominal girth (AG), diagonal body length (DBL), body length (BL), withers height (WH), rump height (RH), hip width (HW), chest width (CW), and body condition score (BCS). Relationships among variables were examined using Pearson correlation analysis and principal component analysis (PCA). Multiple linear regression, Ridge Regression, LASSO regression, Random Forest, and Gradient Boosting algorithms were applied to predict body weight. Additionally, variable importance analysis and the SHAP (SHapley Additive Explanations) approach were utilized to enhance model interpretability. The results demonstrated significant differences between sexes in body weight and the majority of morphometric traits (p < 0.05). Correlation analysis revealed that abdominal girth and heart girth had the strongest associations with body weight. According to PCA results, the first principal component explained the majority of the total variance and represented overall body size. Among the evaluated machine learning models, the Gradient Boosting algorithm achieved the highest prediction performance, with a training R2 of 0.962 and a test R2 of 0.862, together with the lowest prediction errors (RMSE = 1.320 kg and MAE = 1.034 kg). The Random Forest model ranked second, achieving a test R2 of 0.828. Variable importance and SHAP analyses indicated that heart girth and abdominal girth were the most influential features in predicting body weight. In conclusion, morphometric traits can be effectively utilized to predict body weight in Karayaka lambs, and the Gradient Boosting algorithm represents a robust approach offering high accuracy and interpretability.