MLA-C01 Question 4
Select 3A data science team is building a machine learning model using Amazon SageMaker. The dataset resides in an Amazon S3 bucket and contains missing values, categorical variables, and numerical features. The team wants to preprocess the data efficiently and ensure it is ready for training. Which steps should the team take to appropriately prepare the dataset for training?
- A
Use Amazon SageMaker Data Wrangler to handle missing values, encode categorical variables, and scale numerical features.
- B
Write a custom preprocessing script and use SageMaker Processing Jobs to execute it at scale.
- C
Directly train the model on the dataset without handling missing values or encoding categorical variables since SageMaker manages preprocessing automatically.
- D
Use the Amazon SageMaker Feature Store to impute missing values and transform categorical variables into numerical representations.
- E
Leverage AWS Glue to preprocess the data and output a clean dataset to Amazon S3 for model training.
Show answer and explanation
Correct answers: A, B, E
Explanation
Data preparation is essential for building effective machine learning models in SageMaker. Tools like SageMaker Data Wrangler and AWS Glue simplify preprocessing tasks such as handling missing values, encoding categorical variables, and scaling numerical features. Alternatively, custom preprocessing scripts executed using SageMaker Processing Jobs offer flexibility for complex data preparation pipelines. SageMaker itself does not automatically handle preprocessing, and while the Feature Store is useful for feature management, it is not intended for raw data preprocessing.
- A. Correct.
Correct: Amazon SageMaker Data Wrangler provides an intuitive interface and built-in capabilities for handling missing values, encoding categorical variables, and scaling numerical features, making it an effective tool for preprocessing datasets.
- B. Correct.
Correct: Writing a custom preprocessing script and using SageMaker Processing Jobs allows for flexibility and the ability to handle complex preprocessing tasks at scale.
- C. Incorrect.
Incorrect: SageMaker does not handle preprocessing tasks like dealing with missing values or encoding categorical variables automatically. These steps need to be explicitly implemented before training.
- D. Incorrect.
Incorrect: While the Amazon SageMaker Feature Store is useful for storing and serving features, it is not primarily used for preprocessing datasets or transforming raw data.
- E. Correct.
Correct: AWS Glue can preprocess data at scale and output a clean dataset to Amazon S3, which can then be used for training in SageMaker.