Machine Learning
Amazon SageMaker Data Wrangler
Amazon SageMaker Data Wrangler provides a visual and programmatic environment for importing, profiling, transforming, and exporting machine learning data. It helps practitioners prepare structured datasets for customization and evaluation with reusable transformations.
Key points
- Data flows record ordered preparation steps
- Built-in analyses reveal leakage, imbalance, and data quality issues
- Flows can export processing code or jobs into SageMaker workflows
When to use it
- Prepare a preference dataset for model tuning
- Normalize structured evaluation features and inspect class imbalance
Exam tips
- Choose Data Wrangler for interactive ML data preparation and Glue for broad managed data-lake ETL
- Fit transformations only on training data where leakage would invalidate evaluation