Skip to main content

Table 5 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 \(0.8582 \pm 0.0283\) \(\mathbf {0.9306 \pm 0.0216}\) \(0.8004 \pm 0.0244\) \(0.8230 \pm 0.0764\) \(0.6445\pm 0.0720\) \(0.8101 \pm 0.0533\)
CKM-Social \(0.9544 \pm 0.0113\) \(0.8242 \pm 0.0420\) \(0.9685 \pm 0.0138\) \(0.9807 \pm 0.0075\) \(\mathbf {0.9962 \pm 0.0052}\) \(0.9736 \pm 0.0121\)
Congress \(0.9236 \pm 0.0275\) \(0.9362 \pm 0.0217\) \(0.8834 \pm 0.0000\) \(\mathbf {0.9601\pm 0.0000}\) \(0.5370 \pm 0.0953\) \(0.6146 \pm 0.0000\)
DKPol \(0.8332 \pm 0.0163\) \(0.7866 \pm 0.0237\) \(0.5590 \pm 0.0256\) \(0.8424 \pm 0.0212\) \(0.8535 \pm 0.0206\) \(\mathbf {0.8838 \pm 0.0151}\)
Leskovec-Ng \(\mathbf {0.9959 \pm 0.0036}\) \(0.9243 \pm 0.0405\) \(0.9545 \pm 0.0422\) \(0.9797 \pm 0.0022\) \(0.9903 \pm 0.0128\) \(0.9937 \pm 0.0022\)
Starwars \(0.7000 \pm 0.0788\) \(0.7140 \pm 0.0477\) \(0.6319 \pm 0.0771\) \(\mathbf {0.7754 \pm 0.0370}\) \(0.7464 \pm 0.0300\) \(0.6855 \pm 0.0611\)
Terrorist-Noordin \(0.7150 \pm 0.0669\) \(0.7833 \pm 0.0801\) \(0.7000 \pm 0.0939\) \(\mathbf {0.8237 \pm 0.0322}\) \(0.7133 \pm 0.0706\) \(0.6667 \pm 0.0720\)
Terrorist-status \(0.4767 \pm 0.0439\) \(0.5022 \pm 0.0835\) \(0.4915 \pm 0.0554\) \(0.4932 \pm 0.0847\) \(0.4650 \pm 0.0748\) \(\mathbf {0.5034 \pm 0.0809}\)
Vickers \(\mathbf {0.9692 \pm 0.0372}\) \(0.9591 \pm 0.0725\) \(0.8227 \pm 0.1311\) \(0.8909 \pm 0.1563\) \(0.9136\pm 0.0693\) \(0.8227 \pm 0.0756\)
Koumbia-2-mpx \(\mathbf {0.7937 \pm 0.0106}\) \(0.7578 \pm 0.0434\) \(0.6663 \pm 0.0170\) \(0.7520 \pm 0.0064\) \(0.7485 \pm 0.0231\) \(0.7031 \pm 0.0107\)
Koumbia-5-mpx \(\mathbf {0.8404 \pm 0.0132}\) \(0.8314 \pm 0.0289\) \(0.6425 \pm 0.0000\) \(0.8040 \pm 0.0000\) \(0.8077 \pm 0.0128\) \(0.7802 \pm 0.0167\)
Koumbia-10-mpx \(\mathbf {0.8425 \pm 0.0093}\) \(0.8192 \pm 0.0011\) na na \(0.8420 \pm 0.0058\) \(0.8147 \pm 0.0094\)
  1. Bold values refer to the best results on each network