BERT-m7G: A Transformer Architecture Based on BERT and Stacking Ensemble to Identify RNA N7-Methylguanosine Sites from Sequence Information

Table 4

Hyperparameter optimization results of stacking ensemble classifier.

	Classifier	Hyperparameters	Meaning	Search ranges	Optimal values
Base classifiers	LR	C1	The reciprocal of the regularization coefficient	(1, 50)	0.0181
	LightGBM	learning_rate	Learning rate	(0.01, 1.0)	0.2533
		max_depth	Maximum depth of the tree	(1, 50)	12
		max_bin	The max number of bins that feature values will be bucketed in	(10, 100)	84
		boosting_type	Training method	gbdt; goss; dart	gbdt
		num_leaves	Number of leaf nodes	(1, 50)	10
		n_estimators	Number of iterations	(100, 600)	255
	SVM	C2	Regularized constant which determines regularized penalty to estimation errors	(1, 50)	1.1322
		Kernel	Kernel function which uses to realize the nonlinear map from the raw feature space to high-dimensional feature space	Linear; sigmoid; poly; rbf	rbf
Metaclassifier	LR	C3	The reciprocal of the regularization coefficient	(1, 50)	35.5133