MLS-C01 Question 365
Select 3You are building a machine learning model for a retail company to predict customer churn. After training the model, you notice a high variance between the training and validation accuracy. Which of the following strategies could help address this issue?
- A
Increase the size of the training dataset by collecting more customer data.
- B
Reduce the complexity of the model by using fewer layers or parameters.
- C
Add more features to the dataset to provide the model with additional information.
- D
Apply regularization techniques such as L1 or L2 regularization.
- E
Increase the learning rate during training to converge faster.
Show answer and explanation
Correct answers: A, B, D
Explanation
High variance between training and validation accuracy often indicates overfitting. Overfitting occurs when the model performs well on the training data but poorly on unseen data. Strategies such as increasing the training dataset size, reducing model complexity, and applying regularization techniques can help improve generalization and mitigate overfitting. Adding more features or increasing the learning rate are not effective solutions for this specific problem.
- A. Correct.
Increasing the size of the training dataset can help the model generalize better by reducing overfitting, which directly addresses the high variance issue.
- B. Correct.
Reducing the complexity of the model can help prevent overfitting by ensuring the model does not memorize the training data, thus improving validation performance.
- C. Incorrect.
Adding more features may not always help with high variance and could lead to overfitting if not done carefully. This option is not guaranteed to address the issue.
- D. Correct.
Regularization techniques such as L1 or L2 regularization are designed to reduce overfitting by penalizing large weights in the model, which helps in addressing high variance.
- E. Incorrect.
Increasing the learning rate could lead to unstable training and may not address the high variance issue. In fact, it might worsen model performance.