Skip to main content

Table 7 Accuracy (mean and standard deviation over 10 runs) obtained by the proposed methods and competitors, with early-stopping regularization technique

From: Graph convolutional and attention models for entity classification in multilayer networks

Network ML-GAT ML-GCN GrAMME-SG GrAMME-Fusion GAT GCN
Balance \(0.8885 \pm 0.0208\) \(\mathbf {0.8965 \pm 0.0990}\) \(0.7543 \pm 0.0426\) \(0.7982 \pm 0.0666\) \(0.7244\pm 0.0910\) \(0.7881 \pm 0.0270\)
CKM-Social \(0.9623 \pm 0.0088\) \(0.9213 \pm 0.0226\) \(0.9265 \pm 0.0253\) \(0.9479 \pm 0.0192\) \(\mathbf {0.9762 \pm 0.0179}\) \(0.9729 \pm 0.0173\)
Congress \({0.9431 \pm 0.0195}\) \(\mathbf {0.9440 \pm 0.0089}\) \(0.8623 \pm 0.0231\) \(0.9232\pm 0.0303\) \(0.6116 \pm 0.0092\) \(0.6146 \pm 0.0000\)
DKPol \(0.8294 \pm 0.0298\) \(0.7681 \pm 0.0122\) \(0.5910 \pm 0.0433\) \(0.8233 \pm 0.0277\) \(0.7960 \pm 0.0294\) \(\mathbf {0.8621 \pm 0.0321}\)
Leskovec-Ng \(\mathbf {1.0 \pm 0.0000}\) \(0.9570 \pm 0.0178\) \(0.9516 \pm 0.0242\) \(0.9832 \pm 0.0120\) \(0.9792 \pm 0.0202\) \(0.9861 \pm 0.0155\)
Starwars \(\mathbf {0.8174 \pm 0.0183}\) \(0.8152 \pm 0.0185\) \(0.6348 \pm 0.1001\) \(0.6522 \pm 0.1108\) \(0.8087 \pm 0.0171\) \(0.7869 \pm 0.0225\)
Terrorist-Noordin \(0.6500 \pm 0.0745\) \(0.7325 \pm 0.0566\) \(0.7333 \pm 0.1141\) \(\mathbf {0.7179 \pm 0.1216}\) \(0.7175 \pm 0.0426\) \(0.6850 \pm 0.0818\)
Terrorist-status \(0.5700 \pm 0.0483\) \(0.5450 \pm 0.0369\) \(0.5333 \pm 0.0380\) \(0.4974\pm 0.0429\) \(\mathbf {0.6313 \pm 0.0944}\) \(0.5278 \pm 0.0870\)
Vickers \(\mathbf {0.9733 \pm 0.0466}\) \(0.9667 \pm 0.0471\) \(0.8769 \pm 0.1032\) \(0.8974 \pm 0.0544\) \(0.8667\pm 0.0544\) \(0.8000 \pm 0.0943\)
Koumbia-2-mpx \(\mathbf {0.7890 \pm 0.0134}\) \(0.7671 \pm 0.0173\) \({0.6403 \pm 0.0815}\) \({0.7333 \pm 0.0020}\) \(0.6058 \pm 0.0138\) \(0.6846 \pm 0.0139\)
Koumbia-5-mpx \(\mathbf {0.8431 \pm 0.0127}\) \(0.7995 \pm 0.0114\) \({0.5481 \pm 0.0013}\) \({0.8021 \pm 0.0150}\) \(0.7373 \pm 0.0471\) \(0.7715 \pm 0.0141\)
Koumbia-10-mpx \(\mathbf {0.8419 \pm 0.0105}\) \(0.8151 \pm 0.0162\) na na \(0.7825 \pm 0.0612\) \(0.8087 \pm 0.0161\)
  1. Training, validation and testing set sizes correspond to 25%, 25% and 50% of the entities, respectively
  2. Bold values refer to the best results on each network