GraphRag in Action: Revolutionizing Fraud Detection Techniques
summary
Demonstrates a working GraphRAG pipeline that generates Cypher from natural language, runs it against a synthetic financial dataset provided by Neo4j, and returns analyst-friendly explanations; uses a text-to-cypher retriever, an LLM (demo uses Vertex AI with temperature 0) to avoid hallucination, and a Neo4j Aura agent that answers queries like “clients sharing email and phone” and lists matching pairs. Claims feasibility on very large graphs (questions mention 64M nodes up to ~1B nodes) and shows a demo that successfully retrieves and explains suspicious clusters.
AI-written summary · about
chapters (10)
- 0:00Introduction to GraphRAG
- 2:33Understanding Fraud Detection
- 7:12Limitations of Traditional Methods
- 9:03Introducing GraphRAG for Fraud
- 13:35Demo Setup and Dataset
- 15:00Building the GraphRAG Application
- 23:18Creating a Fraud Detection Agent
- 28:07Agent Demo Results
- 29:30Key Takeaways and Summary
- 30:16Q&A Session
related talks
36mGraphRAG Explained 🔥 Building Smarter LLMs
29mCatching Bad Guys using open data and open models for graphs
40mRAG Pipeline Evaluation
01h 17mGraph Database and Retrieval Augmented Generation
36mAgentic GraphRAG: AI’s Logical Edge
30mPerformant GraphRAG Improving Neo4j to Drive Autonomous AI ...
24mUnlocking AI's Hidden Connections With Graphs
01h 40mKnowledge Graphs & GraphRAG: Techniques for Building Effective GenAI Applications