Evaluating machine learning algorithms for accurate prediction of yearling weight in Magra sheep

This study was designed to predict the Twelve-month body weight from live body weights at different age of Magra sheep, reared on Arid Region Campus (ICAR-Central Sheep and Wool Research Institute), Bikaner, Rajasthan using machine learning algorithms. Twelve-month body weight (MWT12) was predicted using several predictors including pedigree records, sex, period, season, dam age at lambing, dam weight at lambing, and early body weight traits such as birth weight (BWT), weaning weight (MWT3), and six-month weight (MWT6). Data of 5,470 Magra sheep born from 1998 to 2023 were used. Comparative analysis was conducted by using Multiple Linear Regression (MLR), Artificial Neural Network (ANN), Bayesian Regression (BR), Support Vector Machine (SVM) and Gradient Boosting Machine (GBM). The algorithms were compared based on goodness of fit criteria including coefficient of determination (R²), root mean square error (RMSE), mean absolute error (MAE) and bias. The results revealed GBM as the top-performing model, achieving an R² value of 0.75 and demonstrating superior predictive accuracy (r = 0.87) with lower values of RMSE, MAE, and bias. Support Vector Machine (SVM) and Artificial Neural Network (ANN) also demonstrated strong predictive capabilities, surpassing traditional approaches such as Multiple Linear Regression (MLR) and Bayesian Regression (BR) in accuracy and overall performance. The present study showed the reliability of advanced machine learning models, particularly GBM, for accurate body weight prediction in sheep breeding and it could be use of potential in genetic selection and flock management in Magra sheep.