Table 4 Comparison of link prediction experimental results of different models.
From: Information flow optimization for adaptive neighbor selection graph embedding
Model | Amazon | DBLP | Retailrocket | ||||||
|---|---|---|---|---|---|---|---|---|---|
ROA | PRA | F1 | ROA | PRA | F1 | ROA | PRA | F1 | |
R-GCN | 0.811 | 0.820 | 0.783 | 0.589 | 0.592 | 0.566 | 0.775 | 0.748 | 0.695 |
MAGNN | 0.958 | 0.949 | 0.915 | 0.602 | 0.699 | 0.684 | 0.847 | 0.847 | 0.872 |
HPN | 0.949 | 0.949 | 0.904 | 0.692 | 0.710 | 0.687 | 0.796 | 0.805 | 0.733 |
PMNE-n | 0.956 | 0.945 | 0.893 | 0.672 | 0.679 | 0.663 | 0.605 | 0.544 | 0.689 |
PMNE-r | 0.884 | 0.890 | 0.796 | 0.637 | 0.640 | 0.629 | 0.531 | 0.503 | 0.613 |
PMNE-c | 0.934 | 0.934 | 0.868 | 0.622 | 0.625 | 0.609 | 0.567 | 0.521 | 0.672 |
MNE | 0.941 | 0.943 | 0.912 | 0.657 | 0.660 | 0.635 | 0.635 | 0.574 | 0.632 |
GATNE | 0.963 | 0.948 | 0.914 | 0.659 | 0.662 | 0.643 | 0.502 | 0.627 | 0.523 |
DMGI | 0.905 | 0.878 | 0.847 | 0.610 | 0.615 | 0.601 | 0.555 | 0.502 | 0.651 |
FAME | 0.959 | 0.951 | 0.900 | 0.742 | 0.750 | 0.733 | 0.646 | 0.634 | 0.636 |
HGTN | 0.960 | 0.950 | 0.897 | 0.787 | 0.832 | 0.775 | 0.836 | 0.854 | 0.778 |
SR-RSC | - | - | - | 0.719 | 0.718 | 0.651 | - | - | - |
Ours | 0.973 | 0.965 | 0.965 | 0.899 | 0.924 | 0.899 | 0.903 | 0.965 | 0.904 |