This AWS Database blog post discusses using Graph Machine Learning (Graph ML) with Amazon Neptune Analytics and GraphStorm to detect fraud in financial transactions. The key highlights include:

• Graph ML can uncover hidden connections and patterns in complex networks that traditional machine learning methods miss
• The solution uses the IEEE CIS fraud dataset with 500,000 transactions, of which 3.5% are fraudulent
• The workflow involves:
  • Exporting data from Neptune Analytics
  • Training a Graph Neural Network (GNN) model using GraphStorm
  • Enriching the graph with embeddings and predictions
  • Performing advanced graph analytics
• The trained model achieved an Area Under the ROC Curve of approximately 0.9
• Techniques like community detection and vector similarity were used to identify suspicious transaction groups

The approach demonstrates how graph machine learning can provide deeper insights into fraud detection by analyzing relationship structures and patterns.