Table 3.

Results for objective 2: ranking performance using different datasets and P -values comparing these models using a paired two-tailed t-test

Dataset	seq-gen	alloc-conc	blind
AUC (standard deviation)
1. Article content	0.769 (0.051)	0.777 (0.034)	0.726 (0.051)
2. PubMed title	0.682 (0.053)	0.690 (0.072)	0675 (0.063)
3. PubMed title and abstract	0.692 (0.037)	0.685 (0.047)	0.694 (0.065)
P -values: 1 comparison of performance using feature sets 1, 2 and 3
1 vs 2	0.001	0.004	<0.001
1 vs 3	<0.001	0.002	0.206
2 vs 3	0.672	0.741	0.497

¹ P -values using two-tailed paired t-test to compare the AUC values across the 10 folds of cross validation.