Table 4.
Hyperparameters of the optimal configuration (lowest validation mean absolute error) for each model for each validation method.
| Model | Validation method | |||
|
|
In-distribution | Out-of-distribution | Cross-validation | |
| Ridge regression | ||||
|
|
α | 0.00 | 0.25 | 0.00 |
| Decision tree regression | ||||
|
|
Depth | 25 | 10 | 5 |
|
|
Minimum sample split | 2 | 5 | 2 |
|
|
Minimum sample leaves | 1 | 1 | 4 |
| Random forest regression | ||||
|
|
Depth | 30 | 15 |
15 |
|
|
Estimators | 150 | 10 | 125 |
|
|
Minimum sample split | 2 | 2 | 2 |
|
|
Minimum sample leaves | 1 | 10 | 10 |
| AdaBoosta regression | ||||
|
|
Estimators | 5 | 5 | 3 |
|
|
Learning rate | 0.1 | 1.0 | 0.1 |
| Support vector regression | ||||
|
|
ε | 0.00 | 0.00 | 0.00 |
|
|
Kernel | Radial | Linear | Linear |
aAdaBoost: adaptive boosting.