Machine Learning Credit Risk Predictive Model

Develop a scikit-learn powered predictive model that assesses credit risk using historical lending data, incorporating advanced feature engineering techniques and handling imbalanced datasets. Implement k-fold cross-validation with stratified sampling, and create a modular pipeline that can integrate different machine learning algorithms (logistic regression, random forest, gradient boosting). Generate comprehensive model performance metrics including precision, recall, F1 score, and ROC-AUC, with a clear visualization of feature importances.