Table 4.

Algorithm’s AUC as a function of feature set for clinical adjudication labels. The AUC is calculated for each algorithm with and without its respective feature selection method. The EMR data that is used is constrained to up to 48 hours before the biomarker measurement and 1 hour after.

Algorithm	All features	EMR	Biomarkers
Logistic Regression	0.78	0.72	0.78
Logistic Regression w/ feature selection	0.79	0.73	0.79
SVM	0.79	0.73	0.78
SVM w/ feature selection	0.81	0.75	0.80
Random Forest	0.77	0.70	0.74
Random Forest w/ feature selection	0.77	0.72	0.77
Adaboost	0.81	0.75	0.79
Adaboost w/ feature selection	0.81	0.76	0.80
Naive Bayes	0.76	0.69	0.77
Naive Bayes w/feature selection	0.80	0.73	0.79