Let's do graph machine learning on Cloud
summary
Uses a Kaggle credit‑card transactions dataset converted to a bipartite NetworkX graph (≈1,000 cards, 700 merchants, ~6,000 frauds in 117,000 transactions) and runs a VertexAI notebook on Google Cloud to produce node embeddings via node2vec and skip‑gram random walks. Edge features are built with Hadamard/average/L1/L2 edge operators, then a RandomForest (100 estimators) classifies fraud; ROC AUC looks satisfactory without heavy tuning. Practical: code, adjacency→embeddings→train pipeline that runs on VertexAI.
AI-written summary · about
chapters (18)
- 0:00Intro
- 1:10Agenda
- 1:52What is a graph
- 2:49Python loves graphs
- 3:31Defining a graph
- 4:26Types of graphs
- 8:05Embeddings
- 9:09Similarity
- 10:23Taxonomy
- 11:02Node to Back
- 12:48Input Graph
- 14:17Use Case
- 16:38Read Data
- 22:29Nodetoback
- 23:15Edge embedding
- 23:55Be careful
- 25:10Conclusion
- 25:34Questions
related talks
12mCloud-Native Friendly Fraud Detection
14mLearn Vertex AI while building a fraud detection system
28mFraud Graph with Amazon Neptune and Machine Learning
28mA Quantum Leap In Fraud Detection Accuracy Using Graph-based Features
56mGraph Analytic Systems
26mAdvancing Financial Fraud Detection with Graph Dbs
45mApplications of Graph in Fraud Detection
27mEnhancing Financial Analytics With Event-Driven Architectures and Graph Neural Networks