Design a Real-Time Fraud Detection System

Design a machine learning system to detect fraudulent transactions in real-time for a large-scale payment platform. Focus on handling imbalanced datasets and implementing adaptive models that can quickly respond to new fraud patterns. Address challenges like minimizing false positives, explaining model decisions, and processing high-volume transaction streams with low latency.