An ML engineer is assembling training data for a binary classification model (labels: Class A, Class B). They want to check whether the classes are balanced and, without removing existing records, bring the dataset into balance if needed. Which approach satisfies this requirement?

The correct approach involves using SageMaker Clarify to detect class imbalance and then applying SMOTE if imbalance is present. SageMaker Clarify is designed to analyze datasets for bias, including class imbalance, by computing metrics like Class Imbalance (CI). A CI greater than 0 indicates an imbalance. Since the requirement specifies "without removing existing records," oversampling techniques are preferred over undersampling. SMOTE (Synthetic Minority Over-sampling Technique) is an effective oversampling method that creates synthetic samples for the minority class, helping to balance the dataset without losing information from the majority class. SageMaker Data Wrangler provides built-in transformations, including SMOTE, to easily apply these techniques. Incorrect options: Applying random undersampling would remove existing records, violating the problem constraint. SageMaker JumpStart is primarily for pre-trained models and solutions, not for generating detailed class imbalance reports or applying data transformations directly. Applying SMOTE when CI equals 0 (meaning no imbalance) is unnecessary and could introduce artificial patterns.