Table 6. Multiple linear regression to predict the number of gap-filling articles appearing in the second time slice in the 10 disease queries only.

As in Table 5 but only for the 2800 gaps in the 10 disease related queries.

Weights of each feature:
0.0301 * sqrt exp query +
−0.0034 * sqrt exp all +
0.0222 * sqrt cooccur all +
0.3411 * sqrt article odds +
1.0526 * pR +
0.0048 * AA score +
−0.2386 * MeSHSim +
−0.4493
Correlation coefficient	0.4712
Kendall’s tau	0.3593
Spearman’s rho	0.4465
Mean absolute error	0.3636
Root mean squared error	0.4923
Relative absolute error	82.8104 %
Root relative squared error	88.1807 %