MLA-C01 Question 176
Select 3You are training a machine learning model to predict customer churn using Amazon SageMaker. Despite hyperparameter tuning, the model's accuracy is below acceptable levels. Which of the following methods can help improve the model's performance?
- A
Collect more labeled training data to increase dataset size
- B
Use feature engineering to create new features or transform existing ones
- C
Switch to a less complex algorithm to reduce overfitting
- D
Apply cross-validation to assess the model's performance
- E
Use a pre-trained model with transfer learning if applicable
Show answer and explanation
Correct answers: A, B, E
Explanation
To improve model performance, it is important to focus on strategies that enhance the quality and quantity of the data, as well as the model's ability to learn from it. Collecting more labeled training data provides the model with more examples to learn from, while feature engineering enables the creation of more meaningful input features. Transfer learning can also be effective, especially when working with large or complex datasets where pre-trained models have already learned useful representations. Other options like cross-validation or switching to a less complex algorithm do not directly address the issue of improving accuracy.
- A. Correct.
Collecting more labeled training data can help improve model performance by providing additional examples for learning patterns and reducing the risk of overfitting.
- B. Correct.
Feature engineering can improve model performance by providing more informative or relevant features, helping the model learn better relationships in the data.
- C. Incorrect.
Switching to a less complex algorithm is typically aimed at reducing overfitting, but in this scenario, the issue is low accuracy, so this approach is not directly helpful.
- D. Incorrect.
Cross-validation is a method for model evaluation and does not directly improve performance; it is useful for assessing how well the model generalizes.
- E. Correct.
Using a pre-trained model with transfer learning can significantly improve performance when applicable, as it leverages knowledge learned from large datasets and related tasks.