Fraud Detection at Floa
summary
Floa explains a real-time fraud pipeline built on Neo4j that ingests heterogeneous product events at about 200 events/min (≈600 DB transactions/min) on a three-hot-node cluster. The pipeline generates Cypher via a query builder to avoid mass overwrites, deduplicates nodes using unique keys, performs entity resolution and community detection to find fraud rings, extracts graph features for ML embeddings, and supports both hot scoring and cold analysis while maintaining low error rates and spare capacity.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 7:08Graph data model
- 16:17Data ingestion pipeline
- 21:35Entity resolution and detection
- 27:33Fraud community demo
- 33:33Deployment and metrics
- 36:30Q&A
related talks
51mReal-time Fraud Detection: Challenges and Solutions
25mReal-Time Fraud Detection at Scale
27mEnhancing Financial Analytics With Event-Driven Architectures and Graph Neural Networks
53mBreaking Credit Card Fraud Rings with Neo4j
12mCloud-Native Friendly Fraud Detection
13mTransforming Financial Services with Real-Time Analytics
10mReal-Time Fraud Prevention at Scale
28mFraud Graph with Amazon Neptune and Machine Learning