Complex Credit Risk Scoring Machine Learning Pipeline

Construct an advanced credit risk assessment machine learning pipeline using PySpark and scikit-learn that integrates multiple data sources for comprehensive borrower evaluation. The model must handle feature engineering from structured (credit history) and unstructured (social media) data, implement ensemble learning techniques, and generate interpretable risk probability scores. Include robust cross-validation, model explainability using SHAP values, and automated reporting of model performance metrics.