MLA-C01 Question 125
Select 2You are tasked with building a machine learning model for a retail company to predict customer churn. The company collects terabytes of real-time customer interaction data daily. You have decided to use Amazon SageMaker to build and deploy the model. Which features of Amazon SageMaker would be most useful in managing this large-scale, real-time data and training the model efficiently?
- A
Using Amazon SageMaker Data Wrangler to preprocess and analyze the real-time data.
- B
Leveraging Amazon SageMaker Feature Store to store and manage features for real-time and batch predictions.
- C
Utilizing Amazon SageMaker Ground Truth to label the real-time data for supervised learning.
- D
Using Amazon SageMaker Training with distributed training to train the model efficiently on large datasets.
- E
Deploying the model using Amazon SageMaker Neo to optimize it for multiple hardware platforms.
Show answer and explanation
Correct answers: B, D
Explanation
Amazon SageMaker Feature Store and SageMaker Training with distributed training are the most relevant solutions for this scenario. Feature Store allows you to manage and serve real-time and batch features effectively, while distributed training ensures that model training on large datasets is efficient and scalable. These features together address the challenges associated with handling terabytes of real-time data and training machine learning models at scale.
- A. Incorrect.
While SageMaker Data Wrangler is useful for data preparation, it is not specifically designed to handle real-time data at the scale described in this scenario.
- B. Correct.
Amazon SageMaker Feature Store is ideal for managing features in real-time and batch workflows, making it suitable for handling a large-scale real-time dataset like this.
- C. Incorrect.
SageMaker Ground Truth is used for labeling datasets and is not relevant here since the question assumes the data is already collected and does not mention a need for labeling.
- D. Correct.
Amazon SageMaker Training with distributed training is highly effective for handling large datasets, as it allows training to be distributed across multiple nodes for efficiency.
- E. Incorrect.
Amazon SageMaker Neo is used to optimize models for deployment on multiple hardware platforms, but this scenario focuses on managing data and efficiently training the model, not deployment optimization.