Table 2.

Performance assessment of the gridSearch compared to the optimizeModel function for model tuning regarding execution time (expressed as HH:MM:SS) and evaluation metric (on the training dataset “Train AUC,” the validation dataset “Val AUC,” given as arithmetic mean across the folds of a 10‐fold cross‐validation) for the four methods implemented in SDMtune

Method	Default model		Genetic algorithm			Grid search
	Train AUC	Val AUC	Train AUC	Val AUC	Time	Train AUC	Val AUC	Time
ANN	0.8600	0.8619	0.9839	0.9590	00:11:44	0.9814	0.9615	05:51:33
BRT	0.9873	0.9750	0.9905	0.9779	00:01:33	0.9892	0.9787	00:29:45
RF	1	0.9724	1	0.9740	00:02:16	1	0.9735	00:48:03
Maxnet	0.8681	0.8561	0.8710	0.8565	00:17:49	0.8702	0.8567	05:01:21

Models were trained using the virtualSp dataset available with the package and 1200 possible hyperparameters' combinations. Presence and background locations were used for the Maxnet method, presence and absence locations for the other methods.