Table A1.

The results of the conducted randomised hyperparameter tuning processes for the classifiers in the article.

Model	Hyperparameter	Search Space	Selected
LR	Regularisation strength	{0, 0.10, 0.20, …, 1}	0.40
	class weight	{0, 1, …, 10}	7
	Maximum number of iterations	{1000, 2000, …, 10,000}	7000
GB	Learning rates	{0.01, 0.02, …, 1}	0.10
	Number of boosting stages	{20, 40, …, 200}	160
	Minimum number of samples required to split an internal node	{1, 2, …, 10}	2
	Minimum number of samples required to be at a leaf node	{{1, 2, …, 10}	6
	Maximum depth of the individual estimators	{1, 2, …, 10}	9
AB	Maximum number of estimators at which boosting is terminated	{10, 20, …, 100}	90
	Learning rates	{0.01, 0.02, …, 1}	1.58
RF	number of trees	{50, 100, …, 500}	20
	Maximum depth of the tree	{1, 2, …, 10}	3
	Minimum number of samples required to split an internal node,	{1, 2, …, 10}	4
	Minimum number of samples required to be at a leaf node	{1, 2, …, 10}	6
	Maximum number of leaf nodes	{1, 2, …, 10}	3
	minimum impurity decrease	{0, 0.001, 0.002, …, 0.010}	0.004
	Cost complexity pruning factor	{0.01, 0.02, …, 0.10}	0.01
	Minimum weighted fraction of the sum total of weights	{0.01, 0.02, …, 0.10}	0.01
SVC	Class weight	{0, 1, …, 10}	6
	Maximum integration	{100, 200, …, 10,000}	2400

Note. AB: AdaBoost; LR: logistic regression; GB: gradient boosting; RF: random forest; SVC: support vector classifier.