FraudShield-AI

An Agentic AI app for Financial Fraud Detection using Graph RAG. An AI-driven fraud detection system leveraging Knowledge Graphs, Graph RAG, ArangoDB, and NetworkX for intelligent, adaptive fraud analysis.

---

Overview

Financial fraud is evolving rapidly, requiring dynamic AI systems that adapt and learn from new patterns. Our Agentic AI uses Graph RAG, LLMs (DeepSeek), and Knowledge Graphs to detect and visualize fraud in financial transactions.

---

Key Features

1. Graph RAG Integration – AI queries knowledge graphs for real-time fraud insights.
2. Agentic AI – Learns & updates fraud patterns dynamically based on interactions.
3. Fraud Detection Algorithms – Uses PageRank, BFS, Shortest Path & Community Detection to flag fraudulent entities.
4. Interactive Graph Visualization – Built with NetworkX & ArangoDB for insightful fraud analysis.
5. Multimodal Querying – Supports text and visual inputs via DeepSeek LLM.

---

Methodology

1. User Inquiry → AI queries financial transactions via Graph RAG.
2. Graph Analysis → Runs Personalized PageRank, BFS, Dijkstra for fraud detection.
3. Dynamic Learning → AI updates fraud patterns using past interactions.
4. Visual Insights → Results are displayed with NetworkX & ArangoDB for fraud tracking.
5. Actionable Alerts → Identifies risky transactions & suggests preventive measures.

---

Tech Stack

• Graph Database: ArangoDB
• Graph Processing: NetworkX
• AI Model: DeepSeek (LLM)
• Backend: Python (FastAPI)
• Frontend: Flutter

---