Studying the potential impact of automated document classification on scheduling a systematic review update

Table 6 Classifier performance on the training and test sets at the closest threshold to a recall of 0.55 on the training set for each topic

		TRAINING SET CROSS-VALIDATION
TOPIC	THRESHOLD	TP	TN	FP	FN	Precision	Recall	F1
ADHD	−0.2466	155	2423	119	130	0.566	0.544	0.555
AEDs	−0.3302	57	1361	75	39	0.432	0.594	0.500
Antiemetics	−0.3550	70	2196	168	63	0.294	0.526	0.377
AtypicalAntipsychotics	−0.1696	375	3544	285	284	0.568	0.569	0.569
MSDrugs	−0.2390	77	1699	75	61	0.507	0.558	0.531
NasalCorticosteroids	−0.3943	8	61	21	5	0.276	0.615	0.381
OveractiveBladder	−0.3461	57	695	128	46	0.308	0.553	0.396
ProtonPumpInhibitors	−0.2532	117	2235	98	87	0.544	0.574	0.558
Sedatives	−0.2617	79	1583	74	66	0.516	0.545	0.530

True positives (TP), true negatives (TN), false positives (FP), false negatives (FN), F1 measure (F1, the harmonic mean of precision and recall).

ISSN: 1472-6947