Gaming Microtransaction Fraud Detection System

Design a machine learning-powered fraud detection system for a gaming platform using Python. Implement an advanced anomaly detection algorithm with TensorFlow that identifies suspicious transaction patterns, including account takeovers, automated bot purchases, and unusual spending behaviors. Create a real-time scoring system that flags potential fraudulent activities with less than 2% false positive rate, integrating historical user behavior, device fingerprinting, and transaction velocity metrics.