TABLE 4.

Performance (SVM and LR) of an active learning method with two separate machine learning (ML) algorithms.

Performance			TP	FP	TN	FN	Recall	Precision	F-score	Precision (Recall = 0.99)
Uncertainty sampling and random negative sampling, two separate ML algorithms, and no validation data update (SVM)	ML1	1st round	178	0	400	22	0.89	1.00	0.94	0.96
		2nd round	132	0	400	68	0.66	1.00	0.80	0.94
	ML2	1st round	190	6	394	10	0.95	0.97	0.96	0.94
		2nd round	150	0	400	50	0.75	1.00	0.86	0.94
Uncertainty sampling and random negative sampling, two separate ML algorithms, and no validation data update (LG)	ML1	1st round	180	3	397	20	0.90	0.98	0.94	0.96
		2nd round	142	2	398	58	0.71	0.99	0.83	0.93
	ML2	1st round	182	0	400	18	0.91	1.00	0.95	0.95
		2nd round	152	2	398	48	0.76	0.99	0.86	0.94