Question

A data scientist is preparing a tabular dataset in Amazon SageMaker Data Wrangler for a customer churn classification project. Before investing time in building a full training pipeline, the team wants a fast estimate of how predictive their current set of engineered features is, and which features contribute most to the prediction, without leaving the Data Wrangler environment. Which action best meets this need?

Accepted Answer

Add a Quick Model analysis to the Data Wrangler flow to train a model on a sample and view a feature importance score for each column

Answer

Export the flow to a SageMaker Processing job and run SHAP analysis with SageMaker Clarify on the full dataset

Answer

Add a Target Leakage analysis to the flow to rank features by their correlation with the target

Answer

Configure a Bias Report analysis in Data Wrangler to measure class imbalance in the target column