Whenever you're ready, walk me through how you'd approach this.
What are the trade-offs between different classification algorithms in a credit risk context?
Discuss how you would compare logistic regression, decision trees, random forests, and gradient boosting for probability-of-default prediction. Address class imbalance, calibration, interpretability, fairness, validation strategy, and deployment constraints. Provide a practical model comparison using the supplied dataset schema rather than assuming a single universally best algorithm.