Table 4 Evaluation performance of all considered ML and PopPK models
Model | RMSE | MAE | ME | \({R^2}\) | MdAPE | MdPE |
|---|---|---|---|---|---|---|
GUH evaluation: a priori | ||||||
GBT new | 34.27 (0.38) | 21.55 (0.25) | − 4.09 (0.00) | 0.58 (0.57) | 17.29% (3.83%) | 0.06% (0.01%) |
GP new | 37.41 (0.43) | 23.54 (0.28) | 2.04 (0.07) | 0.50 (0.46) | 21.39% (4.79%) | − 3.83% (− 0.84%) |
MLP | 38.56 (0.47) | 27.35 (0.34) | 2.58 (0.05) | 0.47 (0.36) | 23.09% (5.29%) | − 5.34% (− 1.26%) |
PopPK | 57.97 (0.64) | 39.67 (0.54) | − 30.27 (− 0.45) | − 0.19 (− 0.21) | 40.79% (11.60%) | 38.33% (11.41%) |
GUH evaluation: a posteriori | ||||||
GBT prev | 32.93 (0.27) | 18.22 (0.19) | − 6.55 (− 0.02) | 0.62 (0.73) | 12.75% (3.09%) | 1.77% (0.43%) |
GP prev | 34.03 (0.28) | 19.41 (0.21) | − 3.83 (− 0.01) | 0.59 (0.71) | 16.48% (3.79%) | − 3.76% (− 0.92%) |
MLP | 37.20 (0.36) | 23.64 (0.26) | − 4.87 (− 0.03) | 0.51 (0.51) | 17.06% (4.14%) | 0.73% (0.17%) |
PopPK | 49.58 (0.43) | 31.28 (0.32) | 4.91 (0.03) | 0.14 (0.32) | 26.09% (6.69%) | − 1.85% (− 0.43%) |
UMCG evaluation: a priori | ||||||
GBT new | 43.92 (0.78) | 38.67 (0.62) | 30.67 (0.58) | 0.36 (− 0.12) | 68.38% (12.89%) | − 68.38% (− 12.89%) |
GP new | 64.99 (0.89) | 55.31 (0.74) | 50.90 (0.72) | − 0.39 (− 0.45) | 84.88% (15.33%) | − 84.88% (− 15.33%) |
MLP | 62.28 (0.85) | 51.47 (0.71) | 38.52 (0.63) | − 0.28 (− 0.33) | 83.09% (14.97%) | − 83.09% (− 14.97%) |
PopPK | 50.46 (0.67) | 31.50 (0.55) | − 23.97 (− 0.30) | 0.16 (0.18) | 39.84% (12.31%) | 33.88 % (9.85%) |
UMCG evaluation: a posteriori | ||||||
GBT prev | 28.12 (0.57) | 21.11(0.40) | 15.01 (0.37) | 0.68 (0.25) | 37.20% (8.46%) | − 37.20% (− 8.46%) |
GP prev | 31.58 (0.57) | 22.73 (0.39) | 18.05 (0.35) | 0.60 (0.26) | 25.15% (6.90%) | − 25.15% (− 6.9%) |
MLP | 30.35 (0.64) | 26.55 (0.50) | 22.45 (0.47) | 0.63 (0.06) | 54.16% (10.48 %) | − 54.16% (− 10.48%) |
PopPK | 25.89 (0.62) | 19.95 (0.45) | 2.15 (− 0.00) | 0.73 (0.13) | 26.69% (7.31%) | 3.31% (0.87%) |