Optimization of Tree-Based Machine Learning Models to Predict the Length of Hospital Stay Using Genetic Algorithm

Table 5

Hyperparameters of tree-based models in the training model and genetic algorithm.

Model	Hyperparameters	Type	Interval	Training model	Genetic algorithm
				Default	Interval	Optimized
Decision tree	max_depth	Int	—	None	(1, 1000)	224
	min_samples_leaf	Int	—	1	(1, 50)	17
	ccp_alpha	Nonnegative float	—	0	(0, 1)	0.9
Random forest	max_features	Int	(1, no. of variables)	1	(1, 12)	6
	n_estimators	Int	(1, inf)	1	(50, 1000)	883
	max_depth	Int	(1, 100)	100	(1, 7)	43
	min_samples_leaf	Int	(1, sample size)	—	(1, 50)	8
	max_samples	Float	(0,1)	1	(0.1, 1)	0.65
XGBoost	learning_rate	Float	(0,1)	0.3	(0.001, 1)	0.4
	n_estimator	Int	(1-Inf)	1000	(50, 1000)	139
	max_depth	Int	(0, inf)	6	(1, 7)	1
	Subsample	Float	(0-1)	1	(0.1, 1)	0.9
	colsample_by_tree	Float	(0-1)	1	(0.1, 1)	0.15
	reg_lambda	Nonnegative float	—	1	(1, 3)	2.8