Databricks Machine Learning Professional Question 222
Single answerYou are monitoring a deployed machine learning model that predicts customer churn. Over time, you observe a consistent drop in model performance. Upon further investigation, you find that while the input features' statistical distributions remain unchanged, the relationship between the features and the target variable has shifted. What type of drift does this scenario represent?
- A
Covariate drift
- B
Concept drift
- C
Prior probability drift
- D
Target label drift
Show answer and explanation
Correct answer: B
Explanation
The scenario describes a situation where the statistical distribution of input features remains unchanged, but the relationship between the features and the target variable (customer churn) has shifted. This is a clear example of concept drift, which directly affects the predictive power of the model.
- A. Incorrect.
Covariate drift refers to changes in the distribution of input features, which is not observed in this scenario as the statistical distributions of the input features remain unchanged.
- B. Correct.
Concept drift occurs when the relationship between input features and the target variable changes over time, which is precisely what is described in this scenario.
- C. Incorrect.
Prior probability drift refers to changes in the distribution of the target variable itself, not the relationship between features and the target variable, so it does not apply here.
- D. Incorrect.
Target label drift is another term often used to describe prior probability drift, which is unrelated to changes in the feature-target relationship seen in this scenario.