Table 3. Discriminative ability of Support Vector Machine (SVM), Random Forest (RF), and L1-Regularised Logistic Regression (RLR) to classify patients likely to benefit from a TC test.

The F₁, F₂, and F₃ scores were obtained from the training data and used to threshold the out of sample test data (indicated by †).

Classifier	AUC	F1 measure			F2 measure			F3 measure
		F1	Se	Sp	F2	Se	Sp	F3	Se	Sp
		score	(%)	(%)	Score	(%)	(%)	Score	(%)	(%)
SVM	0.87	0.54	90	76	0.72	95	72	0.82	95	72
SVM†	0.84		91	69		96	66		96	66
RF	0.84	0.51	82	76	0.68	91	70	0.78	93	67
RF†	0.85		87	71		91	66		91	62
RLR	0.86	0.54	81	80	0.69	86	77	0.79	96	62
RLR†	0.82		75	74		81	71		98	56