Research map / Graph learning / Graph self-supervised learning
BAG: Benchmarking Anomaly Detection on Dynamic Graphs
Fengrui Hua, Yiyan Qi, Zikai Wei, Yuxing Tian, Chengjin Xu, Xiaojun Wu, Jia Li, Jian Guo
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| evaluate on organic anomalies instead of synthetic injected anomalies | all benchmarked DyGAD models · edge-level anomaly detection | AUPRC/AUROC | Hurt |
| use continuous-time (CTDG) models instead of discrete-time (DTDG) models | dynamic GNN anomaly detectors · organic edge anomalies | AUPRC/AUROC | Improved |
| use CTDG instead of DTDG models on synthetic anomalies | dynamic GNN anomaly detectors · synthetic edge anomalies | AUPRC/AUROC | No clear effect |
| use specialized anomaly detection models instead of static GNNs such as SGC and GIN | synthetic edge anomaly detection · synthetic edge anomalies | AUPRC/AUROC | No clear effect |
| decrease the anomaly injection ratio (ablation) | static, DTDG and CTDG models · Bitcoin-OTC synthetic anomalies | AUPRC, F1, Rec@K | Hurt |
| decrease the anomaly injection ratio (ablation) | static, DTDG and CTDG models · Bitcoin-OTC synthetic anomalies | AUROC | No clear effect |
| use attention-based models instead of non-attention counterparts | GNN anomaly detectors without informative node features · synthetic edge anomalies | AUPRC/AUROC | Hurt |
| use too many or too few snapshots for DTDG models (ablation) | DTDG models · synthetic edge anomalies | AUPRC/AUROC | Hurt |
| use CTDG models instead of DTDG or static methods in terms of cost | CTDG models · Bitcoin-OTC (5% anomaly ratio) | runtime and GPU memory | Hurt |
| use tree-based RFGraph and XGBGraph instead of neural models | anomaly detection baselines · Bitcoin-OTC (5% anomaly ratio) | runtime | Improved |
Closest papers by meaning
- Graph Anomaly Detection as Dynamical Transport: Training-Free Scoring via Empirical Bayes (NeurIPS 2026 · similarity 0.91)
- ARC: A Generalist Graph Anomaly Detector with In-Context Learning (NeurIPS 2024 · similarity 0.90)
- A Generalizable Anomaly Detection Method in Dynamic Graphs (AAAI 2025 · similarity 0.89)
- DiffGAD: A Diffusion-based Unsupervised Graph Anomaly Detector (ICLR 2025 · similarity 0.89)
- UniFORM: Towards Unified Framework for Anomaly Detection on Graphs (AAAI 2025 · similarity 0.88)
- Leveraging Diffusion Model as Pseudo-Anomalous Graph Generator for Graph-Level Anomaly Detection (ICML 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
