Graph Anomaly Detection on T-Finance (AUC, F1-Macro)
96.79AUCSignGAD
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SignGADtraining ratio=1%, LLM backbone=GPT-4.12026.05 | 96.79 | 90.42 | |
| TAQ-GADtraining ratio=1%2026.05 | 96.02 | 90.11 | |
| SpaceGNNtraining ratio=1%2026.05 | 95.84 | 91.21 | |
| ConsisGADtraining ratio=1%2026.05 | 95.33 | 90.97 | |
| BWGNNtraining ratio=1%2026.05 | 93.08 | 86.96 | |
| GraphConsistraining ratio=1%2026.05 | 92.61 | 85.37 | |
| DiG-In-GNNtraining ratio=1%2026.05 | 92.53 | 88.51 | |
| GAGAtraining ratio=1%2026.05 | 92.36 | 81.1 | |
| MLPtraining ratio=1%2026.05 | 92.17 | 82.33 | |
| GHRNtraining ratio=1%2026.05 | 91.93 | 80.05 | |
| PC-GNNtraining ratio=1%2026.05 | 91.74 | 86.97 | |
| CARE-GNNtraining ratio=1%2026.05 | 91.45 | 83.68 | |
| GraphSAGEtraining ratio=1%2026.05 | 89.42 | 77.62 | |
| GCNtraining ratio=1%2026.05 | 89.29 | 77.16 | |
| GDNtraining ratio=1%2026.05 | 88.75 | 76.62 | |
| GATtraining ratio=1%2026.05 | 87.4 | 75.49 | |
| GINtraining ratio=1%2026.05 | 81.29 | 65.38 | |
| GATv2training ratio=1%2026.05 | 73.25 | 63.16 |