Comparison of the MARS and XGBoost algorithms for predicting body weight in Kalahari Red goats

Estimating goat body weight from morphometric traits is essential for growth monitoring, health management, and welfare. Accurate weight informs nutrition, clinical interventions, and mating decisions. In this study, the MARS and XGBoost algorithms were used to estimate the live weight of Kalahari Red goats based on body measurements. A dataset containing body length (BL), birth type (single, twin, or triplet), withers height (WH), rump height (RH), and heart girth (HG) information for 200 goats was used in the training and testing of the models. The performances of the models were evaluated with the goodness-of-fit criteria. XGBoost showed a performance of R2 = 0.998, RMSE = 0.023, and MAPE = 0.039 in the training set and R2 = 0.974, RMSE = 0.669, and MAPE = 0.927 in the test set. In addition, MARS achieved R2 = 0.994, RMSE = 0.260, and MAPE = 0.447 in the training set and R2 = 0.968, RMSE = 0.595, and MAPE = 0.771 in the test set. These results demonstrate that, although the R2 values of XGBoost are higher than those of MARS, both algorithms were effective. XGBoost consistently yielded lower errors and slightly higher R2 in estimating the live weight of Kalahari Red goats based on body measurements.