Skip to main content

Table 2 Precision and AUC values obtained for different metrics proposed, with the theoretical bounds of the AUC. We consider dyadic metrics given by Eqs (1)–(4), triadic closure given by the AA metric, Eq. (5), applied to each layer (\(\mathit{AA}_{c}\), \(\mathit{AA}_{r}\), and \(\mathit{AA}_{k}\)), to the aggregated network (\(\mathit{AA}_{a}\)), and the MAA score given by Eq. (6), with coefficients \(\eta _{ck}=0.05\) and \(\eta _{cr}=0.1\) which maximize both AUC and Precision (see Fig. 2). Note that dyadic (MC, NMC, CK, and NCK) and triadic (based on AA) methods use different amount of information, so the theoretical bounds for the AUC are different

From: Prediction of new scientific collaborations through multiplex networks

Method Precision AUC AUC [worst–best]
Random 4.310−4 0.5 [0.50–0.50]
MC 0.025 0.5421 [0.5420–0.5422]
NMC 0.023 0.5422 [0.5420–0.5422]
CK 0.012 0.6648 [0.6082–0.6952]
NCK 0.005 0.6618 [0.6082–0.6952]
\(\mathit{AA}_{c}\) 0.041 0.5635 [0.5633–0.5636]
\(\mathit{AA}_{r}\) 0.017 0.6361 [0.6282–0.6393]
\(\mathit{AA}_{k}\) 0.006 0.6481 [0.0171–0.9951]
\(\mathit{AA}_{a}\) 0.006 0.6495 [0.0147–0.9971]
MAA (all triads) 0.042 0.7620 [0.0147–0.9971]