Databricks Machine Learning Associate Question 230
Select 3A data scientist is building a machine learning model using Databricks and is considering whether to use cross-validation or a simple train-validation split. Which of the following statements correctly describe the benefits or downsides of cross-validation compared to a train-validation split?
- A
Cross-validation provides a more reliable estimate of model performance by reducing variance due to random splits.
- B
Cross-validation is computationally cheaper compared to a train-validation split.
- C
Cross-validation can help detect overfitting or underfitting more effectively than a single train-validation split.
- D
Cross-validation requires more computational resources as it trains the model multiple times.
- E
Cross-validation is always the better choice regardless of the dataset size or complexity.
Show answer and explanation
Correct answers: A, C, D
Explanation
Cross-validation offers significant benefits over a train-validation split, including a more robust estimate of model performance and better detection of overfitting/underfitting. However, it is computationally more expensive as it trains the model multiple times. It is not universally the better choice, and its use depends on the specific requirements and constraints of the task, such as dataset size and available computational resources.
- A. Correct.
Correct: Cross-validation reduces the impact of random splits by training and validating the model on multiple folds, providing a more robust estimate of performance.
- B. Incorrect.
Incorrect: Cross-validation is more computationally expensive than a train-validation split because it trains the model multiple times.
- C. Correct.
Correct: By evaluating the model over multiple folds, cross-validation is better at identifying issues like overfitting or underfitting compared to a single train-validation split.
- D. Correct.
Correct: Cross-validation requires more computational power as the model is trained and validated on multiple splits of the data.
- E. Incorrect.
Incorrect: While cross-validation is often better for performance estimation, it may not always be the best choice for very large datasets or time-sensitive tasks due to its computational cost.