MLA-C01 Question 433
Single answerYou are building a machine learning pipeline using Amazon SageMaker and need to preprocess a large dataset stored in Amazon S3. The dataset requires cleaning, feature engineering, and transformation. The process must scale automatically with the size of the dataset and minimize operational overhead. Which approach should you use?
- A
Use Amazon SageMaker Processing jobs to handle data cleaning, feature engineering, and transformation.
- B
Launch an Amazon EC2 instance, install necessary libraries, and preprocess the data manually.
- C
Use Amazon EMR to preprocess the data and then transfer it to Amazon SageMaker for training.
- D
Use AWS Glue to preprocess the data and directly train models using AWS Glue's ML capabilities.
Show answer and explanation
Correct answer: A
Explanation
Amazon SageMaker Processing jobs are specifically designed to handle preprocessing tasks like data cleaning, feature engineering, and transformation in machine learning workflows. They integrate seamlessly with SageMaker, support large datasets stored in Amazon S3, and scale automatically, making them the most efficient and low-overhead choice for this scenario.
- A. Correct.
This is the correct answer. Amazon SageMaker Processing jobs are designed for preprocessing tasks like data cleaning, feature engineering, and transformation. They integrate well with SageMaker and automatically scale with the size of the dataset, minimizing operational overhead.
- B. Incorrect.
While this approach allows for data preprocessing, it requires significant manual effort to set up, manage, and scale. This does not minimize operational overhead.
- C. Incorrect.
Amazon EMR can preprocess large datasets, but it is better suited for big data processing and analytics use cases. It requires additional steps to integrate with SageMaker, increasing complexity and overhead.
- D. Incorrect.
AWS Glue is a data integration service designed for ETL tasks. While it can preprocess data, it does not natively support training machine learning models, making it a suboptimal choice for this scenario.