MLS-C01 Question 225
Select 2A data science team is building a machine learning model to predict house prices. During the evaluation phase, they notice that the model performs well on the training data but poorly on the test data. What steps should the team take to address this issue?
- A
Increase the size of the training dataset to improve generalization.
- B
Reduce the complexity of the model to prevent overfitting.
- C
Use k-fold cross-validation to better evaluate the model's performance.
- D
Add more features to the model to capture additional patterns in the data.
- E
Apply regularization techniques such as L1 (Lasso) or L2 (Ridge) regularization.
Show answer and explanation
Correct answers: B, E
Explanation
The poor performance on the test data compared to the training data indicates an overfitting issue. Reducing the model's complexity (option 2) and applying regularization techniques (option 5) are effective methods to mitigate overfitting. These approaches help the model generalize better to unseen data without memorizing the training data. While increasing the dataset size (option 1) or adding features (option 4) may help in some cases, they are not guaranteed solutions to this specific problem. K-fold cross-validation (option 3) is useful for evaluation but does not directly address model bias or variance.
- A. Incorrect.
Increasing the size of the training dataset could help reduce overfitting, but it may not always be feasible or effective if the issue is primarily due to model complexity.
- B. Correct.
Reducing the complexity of the model is a standard approach to address overfitting, as it helps to ensure the model does not memorize the training data and can generalize better to unseen data.
- C. Incorrect.
Using k-fold cross-validation is a good evaluation strategy, but it does not directly address the overfitting issue. It only helps provide a more reliable estimate of performance.
- D. Incorrect.
Adding more features can increase the risk of overfitting and may not necessarily improve model performance unless the new features are highly informative.
- E. Correct.
Applying regularization techniques like L1 or L2 can help penalize large weights in the model, reducing overfitting and improving generalization.