Applying Graph Intelligence to Anti-Money Laundering: Business Background, Process, and Case Study

Introduction: The session, organized by DataFun, presents Fabarta’s experience applying graph technology to anti‑money‑laundering (AML) in the financial sector.

Business background : AML faces challenges due to sophisticated, organized laundering activities; traditional rule‑based systems struggle to detect new patterns, prompting the need for graph‑based analysis.

Current AML workflow and pain points : Manual case review, long processing times (20‑50 minutes), reliance on structured SQL queries, and lack of closed‑loop usage of suspicious findings.

Analysis process (five steps): data collection from multiple sources, graph construction (base and feature graphs), model analysis using graph algorithms and machine learning, gang detection, and data circulation for reporting.

Case study : Using an IBM‑provided synthetic transaction dataset, the team demonstrates data collection, feature extraction, graph building, sub‑graph selection, gang discovery, triangle‑count analysis, clustering, and role identification (initiator, intermediary, receiver).

Q&A highlights : discussion on graph topology vs. direction, historical data windows, rule‑based alerts, centrality measures, data integration across institutions, and practical deployment tips.

Conclusion : Graph technology offers superior visualization, integration with other models, and faster, more interpretable AML investigations, though challenges remain in heterogeneous data integration and real‑time querying.