Skip to main content
. Author manuscript; available in PMC: 2024 Jun 13.
Published in final edited form as: J Appl Meteorol Climatol. 2023 Nov;62(11):1539–1572. doi: 10.1175/JAMC-D-22-0142.1

Table 2. Hyperparameter tuning used by each regressor.

Model Parameters dictionary
Linear ‘normalize’: False
Ridge ‘alpha’: 1, ‘normalize’: True, ‘random_state’: 42, ‘solver’: ‘lsqr’, ‘tol’: 0.01
Lasso ‘alpha’: 1, ‘normalize’: False, ‘random_state’: 42, ‘selection’: ‘random’, ‘tol’: 1 × 10 −10
Random forest ‘max_features’: ‘sqrt’, ‘min_samples_leaf’: 11, ‘min_samples_split’: 2, ‘n_estimators’: 400, ‘random_state’: 42
Gradient boosting ‘learning_rate’: 0.2, ‘max_depth’: 3, ‘max_features’: ‘sqrt’, ‘min_samples_leaf’: 10, ‘min_samples_split’: 22, ‘n_estimators’: 200, ‘random_state’: 42, ‘subsample’: 0.2