MLA-C01 Question 289
Select 3You are building a machine learning pipeline for a retail company that wants to predict customer churn. You decide to use Amazon SageMaker for training and deploying the model. During the preprocessing stage, you must handle missing data in the customer dataset. Which of the following methods are both supported by SageMaker and suitable for handling missing data?
- A
Use SageMaker Data Wrangler to impute missing values with the median
- B
Use SageMaker Feature Store to drop rows with missing data
- C
Use the SageMaker Preprocessing Job to fill missing values with a default value
- D
Use SageMaker Processing to implement a custom script for imputing missing values
- E
Use SageMaker Autopilot, which automatically handles missing data during feature engineering
Show answer and explanation
Correct answers: A, D, E
Explanation
Handling missing data is a critical step in machine learning workflows. Amazon SageMaker offers multiple methods to deal with missing data. SageMaker Data Wrangler provides an easy-to-use GUI for common preprocessing tasks, including imputation. SageMaker Processing allows you to write and execute custom scripts for more complex preprocessing needs, and SageMaker Autopilot automates feature engineering, including handling missing values. These tools make it easier to build robust pipelines without manual intervention.
- A. Correct.
Correct. SageMaker Data Wrangler provides an interface to impute missing values, such as replacing them with the median, mean, or a custom value.
- B. Incorrect.
Incorrect. SageMaker Feature Store is primarily used for managing feature data, not for preprocessing tasks like dropping rows with missing data.
- C. Incorrect.
Incorrect. SageMaker does not have a dedicated 'Preprocessing Job' feature for directly filling missing values. Preprocessing is typically done using SageMaker Processing or custom scripts.
- D. Correct.
Correct. SageMaker Processing can execute custom preprocessing scripts (e.g., in Python or R) that handle missing data, including imputation.
- E. Correct.
Correct. SageMaker Autopilot automatically handles missing values as part of its automated feature engineering pipeline.