Whenever you're ready, walk me through how you'd approach this.
What techniques would you use to handle imbalanced datasets?
Discuss data-level methods such as oversampling, undersampling, and synthetic sampling; algorithm-level methods such as class-weighted losses; and decision-level threshold tuning. Explain how you would prevent leakage during resampling, select evaluation metrics, validate the approach, and choose among competing techniques in a practical classification workflow.