MLS-C01 Question 124
Select 3You are building a machine learning model to predict customer churn for a subscription service. The dataset contains categorical features such as 'Subscription Type' and 'Region' and numerical features such as 'Monthly Spend' and 'Tenure (in months)'. While training your model using Amazon SageMaker, you notice that the model's performance is suboptimal. Which of the following actions should you take to improve the model's performance?
- A
One-hot encode the categorical features before training the model.
- B
Normalize the numerical features to ensure they have a mean of 0 and a standard deviation of 1.
- C
Increase the number of training epochs to ensure the model fully learns the patterns in the data.
- D
Perform hyperparameter tuning using SageMaker Automatic Model Tuning.
- E
Remove features with high correlation to avoid multicollinearity issues.
Show answer and explanation
Correct answers: A, B, D
Explanation
Improving model performance often involves addressing both data preprocessing and model optimization. In this scenario, one-hot encoding the categorical features ensures proper representation of categorical data, normalizing the numerical features ensures balanced contributions to the model, and hyperparameter tuning optimizes the model's configuration. While multicollinearity might be a concern for specific types of models, it is not universally necessary to remove correlated features.
- A. Correct.
One-hot encoding categorical features is a common preprocessing step in machine learning. It converts categorical data into a numerical representation that the model can understand, which can improve performance.
- B. Correct.
Normalizing numerical features ensures that all features contribute equally to the model training, preventing features with larger scales from dominating the learning process.
- C. Incorrect.
While increasing the number of training epochs may help in some cases, it is not directly related to addressing data preprocessing issues or optimizing the model's performance.
- D. Correct.
Hyperparameter tuning helps optimize the model's configuration for better performance by systematically searching for the best parameter values.
- E. Incorrect.
Removing features with high correlation (multicollinearity) is usually a concern in linear models, but it is not necessary for most machine learning models like tree-based models or neural networks.