Reproduce “GraphRPM: Risk Pattern Mining on Industrial Large Attributed Graphs” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sheng Tian, Xintan Zeng, Yifei Hu, Baokun Wang, Yongchao Liu, Yue Jin, Changhua Meng, Chuntao Hong and 2 more
Ant Group developed GraphRPM, a scalable framework for risk pattern mining on large attributed financial graphs. It identifies interpretable patterns that distinguish fraudulent from legitimate transactions by combining distributed processing and an enhanced Graph Isomorphism Network, demonstrating high discriminative power and successful deployment in real-world risk control systems.
Key insights
- Focusing on patterns with high discriminative power, measured by a Pattern Risk Score (Rs) (analogous to F1 score), is more effective for risk detection than simply finding frequent patterns.
- A two-staged pattern mining approach, starting with node features then refining with edge features, balances computational efficiency and pattern quality for large attributed graphs.
- Explicitly incorporating edge attributes and using multiple aggregation functions (as in EGIN) is crucial for accurately representing and distinguishing complex patterns in financial transaction graphs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.