Databricks Machine Learning Associate Question 235
Select 4You are training a machine learning model in Databricks to predict customer churn. To ensure the model's performance is robust and generalizable, you decide to perform cross-validation. Which of the following steps are necessary to correctly implement k-fold cross-validation in Databricks?
- A
Split the dataset into k equally-sized folds and rotate the validation fold for each iteration.
- B
Manually train k separate models, each on a different set of training and validation data splits.
- C
Use Databricks' MLflow to automatically log metrics and parameters for each fold during cross-validation.
- D
Combine predictions from all folds to get a final evaluation metric for the model.
- E
Ensure that data leakage is avoided by splitting the data before performing any preprocessing or feature engineering.
Show answer and explanation
Correct answers: A, C, D, E
Explanation
Cross-validation is a crucial step in evaluating machine learning models, as it ensures the model's performance is tested on different subsets of the data. In k-fold cross-validation, the dataset is split into k folds, and the model is trained and validated k times, each time using a different fold as the validation set. Tools like MLflow can automate logging during this process, and care must be taken to avoid data leakage by splitting the data before preprocessing. Combining metrics across folds provides a reliable evaluation metric for the model.
- A. Correct.
Correct. In k-fold cross-validation, the dataset is split into k equally-sized folds, rotating the validation fold for each iteration to ensure that every part of the dataset is used for both training and validation.
- B. Incorrect.
Incorrect. Manually training k separate models is error-prone and inefficient. Libraries like scikit-learn or Databricks' MLflow integration can automate this process, ensuring consistency and reducing manual effort.
- C. Correct.
Correct. MLflow in Databricks can be used to log metrics and parameters for each fold, which helps track model performance across all iterations of cross-validation.
- D. Correct.
Correct. Combining predictions or metrics from all folds provides a final evaluation metric that represents the model's overall performance across the dataset.
- E. Correct.
Correct. To prevent data leakage, ensure the dataset is split before preprocessing or feature engineering. This ensures that information from the validation set does not inadvertently influence the training process.