A data scientist is using AutoML Tables to build a classification model for predicting customer churn. The dataset is highly imbalanced (only 1% churn). Which strategy should they use to handle the class imbalance within AutoML Tables?