MLA-C01 Question 46
Select 3A data science team is building a machine learning model for predicting customer churn using Amazon SageMaker. The dataset contains features with missing values, categorical variables, and numerical variables that are distributed across different scales. As an AWS Certified Machine Learning Engineer, what should you recommend as best practices for preprocessing the data using Amazon SageMaker built-in capabilities?
- A
Use SageMaker Data Wrangler to handle missing values, perform feature scaling, and encode categorical variables.
- B
Use SageMaker Processing Jobs to create a custom script for handling missing values and one-hot encoding categorical variables.
- C
Use SageMaker Feature Store to automatically impute missing values and normalize numerical features.
- D
Rely on SageMaker AutoPilot to preprocess the data automatically during the model creation process.
- E
Directly feed the raw dataset with missing and categorical variables into a SageMaker XGBoost model, as it can handle such data natively.
Show answer and explanation
Correct answers: A, B, D
Explanation
When preprocessing data for machine learning in Amazon SageMaker, it's essential to handle missing values, scale numerical features, and encode categorical variables. SageMaker provides multiple tools for preprocessing, including Data Wrangler for no-code workflows, Processing Jobs for custom scripts, and AutoPilot for automated pipelines. However, relying on raw data without adequate preprocessing can lead to poor model performance or errors.
- A. Correct.
This is correct. SageMaker Data Wrangler provides a no-code interface to handle common preprocessing tasks such as handling missing values, feature scaling, and encoding categorical data.
- B. Correct.
This is correct. SageMaker Processing Jobs allow you to create custom preprocessing scripts to handle specific requirements, including missing values and categorical encoding.
- C. Incorrect.
This is incorrect. SageMaker Feature Store is used to store, retrieve, and manage feature data but does not directly perform preprocessing tasks such as imputing missing values or normalizing features.
- D. Correct.
This is correct. SageMaker AutoPilot performs preprocessing as part of its automated machine learning pipeline, simplifying the process for the user.
- E. Incorrect.
This is incorrect. While XGBoost can handle missing values, it does not natively support categorical variables without preprocessing. Feeding raw data would lead to errors or suboptimal performance.