Table 2.

Hyperparameter values for the random forest and gradient boosted tree algorithms following tuning.

Algorithm	Hyperparameter	Value
Random Forest	mtry: number of features to use for each split in individual trees	20
	num.trees: number of individual trees to build	450
Gradient Boosted Tree	eta: learning rate	.02
	max_depth: complexity of individual trees	12
	nrounds: number of boosting iterations	750
	subsample: proportion of observations to use in each boosting iteration	0.85
	colsample_bytree: proportion of features to use in each boosting iteration	0.57
	colsample_bylevel: proportion of selected features to be used for each split	0.67