Table 2.

Classifier performance on Carevue-only and all data, based on raw training data (unweighted), model class weights (weighted) or synthetic minority oversampling (SMOTE).

Carevue (n = 28,886)
	Unweighted		Weighted		SMOTE
	Precision	Recall	Precision	Recall	Precision	Recall
Braden Score	0.09	0.5
LR	0.1	0.005	0.09	0.67	0.12	0.49
Elastic Net	0.11	0.005	0.09	0.67	0.13	0.48
SVM	NA	0	NA	0	0.12	0.48
RF	0.33	0.02	0.3	0.02	0.18	0.17
GBM	0.11	0.18	0.04	0.91	0.16	0.36
Neural net	0.33	0.01	0.09	0.7	0.11	0.52
All Data (n = 50,851)
LR	0.67	0.006	0.09	0.71	0.12	0.46
Elastic Net	0.67	0.006	0.09	0.7	0.12	0.46
SVM	NA	0	NA	0	0.12	0.44
RF	NA	0	0.33	0.006	0.18	0.28
GBM	0.11	0.18	0.04	0.94	0.17	0.38
Neural net	0.06	0.006	0.09	0.7	0.11	0.49