Whenever you're ready, walk me through how you'd approach this.
What techniques would you use to handle class imbalance in a dataset?
Discuss how you would choose among class weighting, over-sampling, under-sampling, synthetic sampling, threshold tuning, and anomaly-detection approaches. Explain how you would prevent data leakage and select evaluation metrics that reflect minority-class performance. Provide a practical scikit-learn implementation for a binary classification dataset with numerical and categorical features, missing values, and an unspecified positive-class prevalence.