Graph-Based
Fraud Detection

Turn transactions into a network.

Instead of treating every transaction as an isolated row, the project models financial activity as a graph so relationships, cycles and suspicious clusters become visible.

The pipeline combines graph traversal techniques — including cycle detection and connected components — with Isolation Forest anomaly detection. A Flask layer exposes the analysis through a dashboard so the result is not just a model score, but an explorable system.