Machine Learning Credit Risk Prediction Pipeline

Architect a comprehensive machine learning pipeline for credit risk prediction that can integrate multiple data sources, including traditional credit history, alternative data streams, and real-time economic indicators. Design a modular system that supports model retraining, feature engineering, and can dynamically adapt to changing economic conditions. Include comprehensive logging, model interpretability features, and a mechanism for tracking prediction confidence intervals.