Table 1 The optimal hyperparameters for different regressors.
From: Prediction of loess collapsibility coefficient using bayesian optimized random forest model
Regressor | Search space | Best parameter |
|---|---|---|
DecisionTree | max_depth: (1, 20), min_samples_split: (2, 20), min_samples_leaf: (1, 20) | 8 2 1 |
Ridge | alpha: (0.001, 100, ‘log-uniform’) | 4.3 |
RandomForest | n_estimators: (10, 200), max_depth: (3, 15), min_samples_leaf: (1, 10), min_samples_split: (2, 20) | 152 13 1 2 |
SVR | C: (1e-3, 100, ‘log-uniform’), epsilon: (0.01, 1, ‘uniform’), gamma: (1e-3, 1, ‘log-uniform’), kernel: [‘linear’, ‘poly’, ‘rbf’,‘sigmoid’] | 100 0.026 0.503 rbf |
LGBM | num_leaves: (10, 100), learning_rate: (0.001, 100, ‘log-uniform’), n_estimators: (100, 1000), min_data_in_leaf: (10, 100), max_depth: (3, 10) | 20 0.03 629 10 7 |
xGBoost | learning_rate: (0.001, 0.1), n_estimators: (100, 1000), max_depth: (3, 10), min_child_weight: (1,9) | 0.1, 470 10 9 |
