MLA-C01 Question 437
Single answerA data scientist is working on a machine learning project using Amazon SageMaker. They are tasked with training a regression model to predict housing prices. During the training process, they notice the model's performance is poor, and suspect the issue lies with improper feature scaling. What is the BEST way to address this issue using SageMaker's built-in capabilities?
- A
Use SageMaker's built-in Linear Learner algorithm, which automatically handles feature scaling during training.
- B
Manually normalize the dataset using a custom preprocessing script and upload the normalized data to an S3 bucket for training.
- C
Enable automatic scaling in Amazon SageMaker by configuring the training instance's hyperparameters.
- D
Use SageMaker's built-in Data Wrangler tool to apply feature scaling before passing the dataset to the training job.
Show answer and explanation
Correct answer: A
Explanation
Feature scaling is critical for many machine learning models to ensure that features with different scales do not dominate the training process. SageMaker's Linear Learner algorithm simplifies this process by performing feature scaling automatically, eliminating the need for manual preprocessing or additional tools. This makes it the best choice for addressing the issue in this scenario.
- A. Correct.
Correct: SageMaker's Linear Learner algorithm automatically normalizes features as part of its training process, making it the best and simplest option in this scenario.
- B. Incorrect.
Incorrect: While manually normalizing the dataset is a valid approach, it is more labor-intensive and not necessary when using the Linear Learner algorithm, which handles feature scaling automatically.
- C. Incorrect.
Incorrect: Amazon SageMaker's instance configuration does not include an option to enable automatic feature scaling through hyperparameters. This is not a valid solution.
- D. Incorrect.
Incorrect: While SageMaker Data Wrangler can be used for preprocessing, using it is unnecessary here, as the Linear Learner algorithm already performs feature scaling internally.