Table 1.

Overview of the hyperparameters that can be tuned per statistical method and underlying package

Method	R package	Hyperparameters	Default value
ANN	nnet	Size of hidden layer	–
	(Venables & Ripley, 2002)	Weight decay	0
		Initial random weights	0.7
		Number of iterations	100
BRT	gbm	Number of trees	100
	(Greenwell, Boehmke, & Cunningham, 2019)	Interaction depth	1
		Shrinkage	0.1
		Bag fraction	0.5
ME	dismo	Feature class combinations	lqph a
	(Hijmans, Phillips, Leathwick, & Elith, 2017)	Regularization multiplier	1
		Number of iterations	500
	maxnet	Feature class combinations	lqph
	(Phillips, 2017b)	Regularization multiplier	1
RF	randomForest	Number of randomly sampled variables	floor(sqrt(#variables))
	(Liaw & Wiener, 2002)	Number of trees	500
		Minimum size of terminal nodes	1

The meaning of each hyperparameter can be found in the respective package documentation and default values, when available, are provided in the last column.

^a(l) linear, (q) quadratic, (p) product, and (h) hinge.