Databricks Machine Learning Associate Question 581
Select 2A data scientist is working on a machine learning project and is deciding between using cross-validation and a simple train-validation split for model evaluation. Which of the following are benefits of using cross-validation compared to a train-validation split?
- A
It provides a more robust estimate of model performance by averaging results across multiple folds.
- B
It is computationally faster than a train-validation split because it uses the same data repeatedly.
- C
It reduces the risk of overfitting by ensuring the model is tested on multiple subsets of the data.
- D
It allows for automated hyperparameter tuning without the need for a separate validation set.
- E
It eliminates the need for a test set in the machine learning workflow.
Show answer and explanation
Correct answers: A, C
Explanation
Cross-validation is a robust technique for model evaluation because it uses multiple training and validation splits, providing a more reliable estimate of model performance compared to a single train-validation split. It also reduces overfitting by ensuring the model is tested on diverse subsets of the data. However, cross-validation is computationally expensive, and it does not replace the need for a separate test set or validation set in certain workflows.
- A. Correct.
Correct: Cross-validation provides a more robust estimate of model performance since it evaluates the model on multiple training and validation splits, reducing the variance introduced by a single train-validation split.
- B. Incorrect.
Incorrect: Cross-validation is generally more computationally expensive than a train-validation split because the model is trained multiple times (once for each fold).
- C. Correct.
Correct: By testing the model on multiple subsets of the data, cross-validation reduces the likelihood of overfitting to any specific subset of the data.
- D. Incorrect.
Incorrect: While cross-validation is useful for model evaluation, it does not replace the need for a separate validation set during automated hyperparameter tuning.
- E. Incorrect.
Incorrect: Cross-validation does not eliminate the need for a test set, which is required for final, unbiased evaluation of the model performance.