Table 5.

Performance of classification models on the test dataset; outcome – TTF3.

Outcome	Model	Features	Class	N. class	Precision	Recall	F1	Acc.	AUC
TTF3 Class 0 (TTF<3m) 213 patients Class 1 (TTF3≥6m) 267 patients	CB	27	0	23	0.79	0.83	0.81	0.81	0.81
			1	25	0.83	0.80	0.82
	LR	9	0	23	0.76	0.70	0.73	0.75	0.79
			1	25	0.74	0.80	0.78
	NN	10	0	23	0.74	0.74	0.74	0.75	0.78
			1	25	0.76	0.76	0.76
	RF	5	0	23	0.57	0.57	0.57	0.58	0.65
			1	25	0.60	0.60	0.60
	SVM	11	0	23	0.53	0.78	0.63	0.56	0.51
			1	25	0.64	0.36	0.46

ACC, testing accuracy; AUC, area under the curve; TTF3, 3-months’ time to treatment failure; N. class, number of patients in both classes for a test set. Bold is the best performing model for this outcome.