Skip to main content

Table 6 Accuracy (mean and standard deviation over 10 runs) obtained by the proposed methods and competitors

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

Network ML-GAT ML-GCN GrAMME-SG GrAMME-Fusion GAT GCN
Balance \(0.7265 \pm 0.0270\) \(\mathbf {0.8164 \pm 0.0139}\) \(0.6795 \pm 0.0377\) \(0.7188 \pm 0.0483\) \(0.5496\pm 0.0852\) \(0.7670 \pm 0.0317\)
CKM-Social \(0.8652 \pm 0.0193\) \(0.7819 \pm 0.0213\) \(0.8026 \pm 0.0649\) \(0.8638 \pm 0.0596\) \(\mathbf {0.9815 \pm 0.0189}\) \(0.9427 \pm 0.0281\)
Congress \(0.9262 \pm 0.0212\) \(\mathbf {0.9342 \pm 0.0065}\) \(0.8814 \pm 0.0000\) \(0.8879\pm 0.0000\) \(0.5398 \pm 0.0649\) \(0.6146 \pm 0.0000\)
DKPol \(0.6809 \pm 0.0510\) \(0.5377 \pm 0.2600\) \(0.3679 \pm 0.0377\) \(0.5373 \pm 0.0513\) \(0.7384 \pm 0.0577\) \(\mathbf {0.7969 \pm 0.0394}\)
Leskovec-Ng \(\mathbf {0.9872 \pm 0.0053}\) \(0.8750 \pm 0.0454\) \(0.7873 \pm 0.0960\) \(0.8746 \pm 0.0749\) \(0.9867 \pm 0.0191\) \(0.9867 \pm 0.0146\)
Starwars \(0.7465 \pm 0.0455\) \(\mathbf {0.7767 \pm 0.0741}\) \(0.6885 \pm 0.1278\) \(0.6828 \pm 0.1861\) \(0.6023 \pm 0.0757\) \(0.6151 \pm 0.1531\)
Terrorist-Noordin \(0.6568 \pm 0.0428\) \(0.7284 \pm 0.0149\) \(0.6536 \pm 0.0947\) \(\mathbf {0.7427 \pm 0.0860}\) \(0.6649 \pm 0.0774\) \(0.6324 \pm 0.0734\)
Terrorist-status \(0.4648 \pm 0.0377\) \(0.4767\pm 0.0328\) \(\mathbf {0.5460 \pm 0.0790}\) \(0.5360 \pm 0.0917\) \(0.4575 \pm 0.0810\) \(0.4247 \pm 0.0961\)
Vickers \(\mathbf {0.9296 \pm 0.0729}\) \(0.9074 \pm 0.0400\) \(0.5750 \pm 0.1745\) \(0.5643 \pm 0.1355\) \(0.7296\pm 0.1063\) \(0.7851 \pm 0.0716\)
Koumbia-2-mpx \(\mathbf {0.6551 \pm 0.0121}\) \(0.6405 \pm 0.0448\) \(0.5999 \pm 0.0188\) \({0.6404 \pm 0.0064}\) \(0.5865 \pm 0.0182\) \(0.5835 \pm 0.0187\)
Koumbia-5-mpx \(\mathbf {0.7457 \pm 0.0286}\) \(0.6677 \pm 0.0194\) \(0.5455 \pm 0.0156\) \(0.7147 \pm 0.0120\) \(0.6698 \pm 0.0162\) \(0.6426 \pm 0.0339\)
Koumbia-10-mpx \(\mathbf {0.7674 \pm 0.0596}\) \(0.6265 \pm 0.0252\) na na \(0.7278 \pm 0.0179\) \(0.6694 \pm 0.0156\)
  1. Training and testing set sizes correspond to 5% and 95% of the entities, respectively
  2. Bold values refer to the best results on each network