Table 3 Implemented algorithms and their...

Table 3

Implemented algorithms and their hyperparameter values

No.	Models	Parameters	Values	Standard range
1	RR	alpha	10.0	0.1–100 (log scale)
		fit_intercept	True	Boolean (true/false)
		solver	Sparse_cg	{“auto”, “svd”, “cholesky”, “lsqr”, “sparse_cg”, “sag”, “saga”}
2	Lasso	alpha	0.1	0.0001–1 (log scale)
		fit_intercept	True	Boolean (true/false)
		selection	random	{“cyclic”, “random”}
3	SVM	c	100	0.1–1000 (log scale)
		gamma	scale	{“scale”, “auto”} or float (0.001–10)
		kernel	rbf	{“linear”, “poly”, “rbf”, “sigmoid”}
4	KNN	algorithm	brute	{“auto”, “ball_tree”, “kd_tree”, “brute”}
		leaf_size	10	1–100
		metric	Manhattan	{“euclidean”, “manhattan”, “minkowski”}
		n_neighbors	3	1–20
		p	1	1 (manhattan), 2 (euclidean)
		weights	distance	{“uniform”, “distance”}
5	DT	criterion	squared_error	{“squared_error”, “friedman_mse”, “absolute_error”}
		max_depth	20	1–100
		max_features	sqrt	{“sqrt”, “log2”, None} or int/float
		max_leaf_nodes	None	2–infinity or None
		min_sample_leaf	1	1–20
		min_sample_split	2	2–20
		random_state	42	Fixed seed
		splitter	best	{“best”, “random”}
6	RF	max_depth	30	1–100
		max_features	sqrt	{“sqrt”, “log2”, None} or int/float
		min_sample_leaf	1	1–20
		min_sample_split	2	2–20
		n_estimator	200	10–1,000
7	AB	learning_rate	1.0	0.01–1
		n_estimator	50	10–1,000
		random_state	72	Fixed seed
8	XGB	colsample_bytree	0.4	0.1–1
		gamma	0.1	0–infinity (typically 0–5)
		learning_rate	0.15	0.001–0.3
		max_depth	6	1–20
		min_child_weight	1	0–infinity (typically 1–10)
9	CB	learning_rate	0.3	0.001–0.3
		n_estimator	100	10–1,000
		random_state	50	Fixed seed
10	LGBM	learning_rate	0.1	0.001–0.3
		max_depth	5	1–20
		min_data_in_leaf	20	1–100
		n_estimator	450	10–1,000

No.	Models	Parameters	Values	Standard range
1	RR	alpha	10.0	0.1–100 (log scale)
		fit_intercept	True	Boolean (true/false)
		solver	Sparse_cg	{“auto”, “svd”, “cholesky”, “lsqr”, “sparse_cg”, “sag”, “saga”}
2	Lasso	alpha	0.1	0.0001–1 (log scale)
		fit_intercept	True	Boolean (true/false)
		selection	random	{“cyclic”, “random”}
3	SVM	c	100	0.1–1000 (log scale)
		gamma	scale	{“scale”, “auto”} or float (0.001–10)
		kernel	rbf	{“linear”, “poly”, “rbf”, “sigmoid”}
4	KNN	algorithm	brute	{“auto”, “ball_tree”, “kd_tree”, “brute”}
		leaf_size	10	1–100
		metric	Manhattan	{“euclidean”, “manhattan”, “minkowski”}
		n_neighbors	3	1–20
		p	1	1 (manhattan), 2 (euclidean)
		weights	distance	{“uniform”, “distance”}
5	DT	criterion	squared_error	{“squared_error”, “friedman_mse”, “absolute_error”}
		max_depth	20	1–100
		max_features	sqrt	{“sqrt”, “log2”, None} or int/float
		max_leaf_nodes	None	2–infinity or None
		min_sample_leaf	1	1–20
		min_sample_split	2	2–20
		random_state	42	Fixed seed
		splitter	best	{“best”, “random”}
6	RF	max_depth	30	1–100
		max_features	sqrt	{“sqrt”, “log2”, None} or int/float
		min_sample_leaf	1	1–20
		min_sample_split	2	2–20
		n_estimator	200	10–1,000
7	AB	learning_rate	1.0	0.01–1
		n_estimator	50	10–1,000
		random_state	72	Fixed seed
8	XGB	colsample_bytree	0.4	0.1–1
		gamma	0.1	0–infinity (typically 0–5)
		learning_rate	0.15	0.001–0.3
		max_depth	6	1–20
		min_child_weight	1	0–infinity (typically 1–10)
9	CB	learning_rate	0.3	0.001–0.3
		n_estimator	100	10–1,000
		random_state	50	Fixed seed
10	LGBM	learning_rate	0.1	0.001–0.3
		max_depth	5	1–20
		min_data_in_leaf	20	1–100
		n_estimator	450	10–1,000

Source(s): Authors’ own work

[ViewLarge]

Sharing Unavailable