User Behavior Tutorial

Abstraction: KDD 2015 tutorial connecting graph analysis to user behavior modeling and fraud detection

Key points:

• Tutorial by Alex Beutel (CMU), presented at KDD 2015
• Central question: how to model normal user preferences and detect when fraud/spam/anomalies distort those models
• Three core techniques applied to graphs: subgraph analysis, label propagation, and latent factor models
• Covers static, evolving, and attributed graph settings
• Shows how each method's outputs are affected by fraudsters and how the same techniques can be repurposed for fraud detection
• Bridges two research threads: graph-based user behavior modeling and state-of-the-art fraud/spam detection research

Connections: Carnegie Mellon University · User Behavior Modeling · Fraud Detection · Graph Analysis · Anomaly Detection

Source: http://www.cs.cmu.edu/~abeutel/kdd2015_tutorial/