Figure 4.

Prediction accuracy of cross-validation across four fruit-related traits by different genomic prediction methods. BLUP values for each trait were applied to 10 different genomic prediction models. Boxes indicate the range of upper quartile and lower quartile, and the bar in the box is the median. Whiskers from upper quartile to maximum and lower quartile to minimum were vertical lines. Black spots were outliers. Reproducing Kernel Hilbert Space (RKHS) was the most effective model for fruit length (FL). Random forest showed high performance for fruit width (FWd). Among the four traits, FWd had the highest accuracy with the highest heritability. Random forest showed the highest accuracy for fruit weight (FWg). Prediction accuracy of the Bayes C method was unstable. Random forest showed the highest performance for pericarp thickness (PT).