Databricks Machine Learning Professional Question 225
Single answerYou are monitoring the performance of a deployed machine learning model that predicts customer churn. Over time, you notice that while the model's features (such as customer age, account balance, and subscription type) exhibit stability in their distributions, the proportion of customers who churn has significantly increased compared to the training data. Which of the following best explains this situation?
- A
This is an example of label drift, as the distribution of the target variable has changed.
- B
This is an example of feature drift, as the distribution of input features has changed.
- C
This is an example of label drift, as the features remain stable while the relationships between features and the target variable have changed.
- D
This is an example of feature drift, as the target variable’s distribution impacts the features’ behavior.
Show answer and explanation
Correct answer: A
Explanation
Label drift refers to changes in the distribution of the target variable over time, while feature drift relates to changes in the distributions of input features. In this scenario, the input features were stable, but the target variable (customer churn) exhibited significant distributional changes, indicating label drift. Identifying and addressing drift types is critical for maintaining the reliability of machine learning models in production.
- A. Correct.
Correct. Label drift occurs when the distribution of the target variable (in this case, customer churn) changes over time, even if the input features remain stable.
- B. Incorrect.
Incorrect. Feature drift refers to changes in the distribution of input features, but in this scenario, the features are stable.
- C. Incorrect.
Incorrect. While label drift refers to changes in the target variable’s distribution, it does not imply changes in the relationships between features and the target variable. This statement is misleading.
- D. Incorrect.
Incorrect. Feature drift is unrelated to the target variable’s distribution; it specifically refers to changes in the input features’ distribution.