Databricks Machine Learning Professional Question 221
Single answerYou are working on a machine learning model to predict customer churn for a subscription-based service. After deploying the model, you notice that the model's performance has degraded over time. Upon investigation, you find that the features related to customer behavior (e.g., login frequency, session duration) have shifted in their statistical distribution compared to the training data. Which type of drift is most likely responsible for this change?
- A
Concept drift
- B
Covariate drift
- C
Prediction drift
- D
Feature drift
Show answer and explanation
Correct answer: B
Explanation
The problem described involves a shift in the statistical distribution of the input features (customer behavior), which aligns with the definition of covariate drift. Understanding and identifying different types of drift is crucial for maintaining model performance over time in a production environment.
- A. Incorrect.
Concept drift refers to a change in the relationship between the input features and the target variable. While it can impact model performance, the described scenario mentions a shift in the feature distribution, not the relationship itself.
- B. Correct.
Covariate drift refers to a change in the distribution of the input features while the relationship between the features and the target variable remains stable. This matches the scenario, as the statistical distribution of customer behavior features has shifted.
- C. Incorrect.
Prediction drift refers to a change in the distribution of the model's predictions over time. This is not directly relevant in the given scenario, which focuses on input feature distribution.
- D. Incorrect.
Feature drift is not a commonly used term in data science or machine learning, and it does not accurately describe the issue in the scenario.