Skip to main content
. 2023 Nov 7;101:skad376. doi: 10.1093/jas/skad376

Table 5.

Validation results from regression models using the 7:3 random split

BCS PLS DL GBM
Exact 0.5-unit DEV Exact 0.5-unit DEV Exact 0.5-unit DEV
1.5 33 (19-52) 91 (80-100) 16 (6-29) 89 (72-100) 6 (0-18) 84 (76-94)
2.0 49 (46-52) 97 (94-99) 50 (45-54) 97 (95-99) 53 (47-58) 97 (96-99)
2.5 60 (56-69) 98 (96-99) 67 (61-73) 98 (97-99) 62 (57-66) 98 (96-99)
3.0 55 (51-58) 98 (97-100) 52 (44-60) 98 (97-99) 51 (41-57) 97 (95-99)
3.5 45 (36-51) 94 (91-99) 41 (34-48) 91 (86-98) 39 (33-45) 90 (83-94)
4.0 23 (10-35) 86 (80-96) 23 (7-29) 78 (71-83) 23 (10-31) 75 (62-91)
4.5 9 (0-20) 65 (42-78) 9 (0-20) 64 (40-78) 2 (0-20) 47 (20-73)
WAvg 51.2 96.1 52.0 95.5 50.4 94.3
R 2 0.67 (0.65-0.68) 0.66 (0.64-0.68) 0.63 (0.61-0.66)
RMSE 0.31 (0.29-0.33) 0.29 (0.26-0.32) 0.30 (0.28-0.32)

Accuracy of the BCS categories is presented in percentage and parenthesis represents the range among replicates. PLS = Partial Least Square, DL = DeepLearning, GBM = Gradient Boosting Machine, BCS = Body Condition Score, R2 = R-square, RMSE = Root Mean Square Error, Exact = exact score, DEV = deviation, WAvg = weighted average by frequency.