Animal Molecular Breeding 2024, Vol.14, No.1, 95-105 http://animalscipublisher.com/index.php/amb 98 Srivastava et al. (2021) demonstrated the results of predictive correlation and mean squared error (MSE) for four carcass traits of Korean cattle using GBLUP and three machine learning methods (RF, XGB, SVM) (Figure 2). XGB showed the highest predictive correlation for CWT and MS, followed by GBLUP, SVM, and RF. GBLUP exhibited the greatest predictive correlation for BFT and EMA, followed by SVM, RF, and XGB. In terms of measuring prediction performance, although predictive correlation is a commonly used straightforward approach, MSE is a preferable parameter considering prediction bias and variance. Among all traits, GBLUP performed best in terms of MSE, while among the machine learning methods, XGB performed optimally for CWT and MS, and SVM was best for BFT and EMA. Figure 2 Predictive correlation (red color) and mean squared error (blue color) of prediction obtained using different statistical methods for carcass weight (CWT), marbling score (MS), backfat thickness (BFT), and eye muscle area (EMA) (Photo credit: Srivastava et al., 2021) Image caption: RF: random forest; XGB: extreme gradient boosting; SVM: support vector machine; GBLUP: genomic best linear unbiased prediction (Adopted from Srivastava et al., 2021)
RkJQdWJsaXNoZXIy MjQ4ODY0NQ==