Table 2. Hyperparameters for the XGBoost model.

Hyperparameter	Meaning	Range of values	Optimal results
			(BO-GP)	(BO-RF)	(RS)
’n_estimators’	Number of trees	100–1000	1000	912	823
’max_depth’	Maximum depth of each tree	3–9	3	3	3
’learning_rate’	Learning rate of stages	0.05–0.30	0.1399	0.1194	0.109
’booster’	Booster method	‘gbtree’, ‘dart’	‘dart’	‘gbtree’	‘dart’
’gamma’	The minimum loss to create a tree’s nodes	0.01–0.50	0.5	0.485	0.3077
’subsample’	The subsampling ratio in the training set	0.60–0.90	0.6	0.697	0.727
’colsample_bytree’	Specifies the proportion of columns to be subsampled	0.60–0.90	0.9	0.747	0.799
’reg_lambda’	Weights used in L2 regularization	1–50	22	3	8

The optimal results show that all methods select a fairly large number of trees (from 823–1000 trees). The maximum depth of all trees is only 3 while the remaining hyperparameters are chosen differently depending on each algorithm.