MLA-C01 Question 44
Select 3You are working on a machine learning project to predict customer churn using data stored in an Amazon S3 bucket. The dataset contains missing values, categorical features, and numerical features. You need to prepare the data for training using AWS services. Which of the following actions should you take to perform effective data transformation and feature engineering?
- A
Use AWS Glue DataBrew to handle missing values and encode categorical variables.
- B
Use Amazon QuickSight to create visualizations for data exploration before feature engineering.
- C
Normalize numerical features using Amazon SageMaker Processing with a custom Python script.
- D
Replace missing values with default values and convert categorical variables into numerical format directly within the Amazon S3 bucket.
- E
Leverage Amazon SageMaker Feature Store to store and retrieve engineered features for consistent use across multiple models.
Show answer and explanation
Correct answers: A, C, E
Explanation
Effective data transformation and feature engineering require the use of appropriate AWS services such as AWS Glue DataBrew for preprocessing, Amazon SageMaker Processing for custom transformations like normalization, and Amazon SageMaker Feature Store for managing engineered features. While Amazon QuickSight is useful for exploratory data analysis, it is not designed for data transformation. Similarly, S3 is a storage solution and does not support direct feature engineering operations.
- A. Correct.
AWS Glue DataBrew is a data preparation tool that can assist in handling missing values and encoding categorical variables, making it a suitable choice for feature engineering.
- B. Incorrect.
Although Amazon QuickSight is a powerful visualization tool for exploring data, it is not typically used for performing feature engineering or data transformation tasks.
- C. Correct.
Amazon SageMaker Processing allows you to run preprocessing scripts, such as normalizing numerical features, using custom Python code, making it a valid option.
- D. Incorrect.
Data stored in an Amazon S3 bucket cannot be directly transformed. Missing value imputation and categorical encoding require dedicated tools or processing jobs.
- E. Correct.
Amazon SageMaker Feature Store is designed to enable consistent storage and retrieval of engineered features, which is crucial for maintaining consistency across different models and workflows.