Databricks Machine Learning Professional Question 223
Single answerYou are monitoring a deployed machine learning model that predicts customer churn for a subscription-based service. Over the last three months, your team has observed that while the feature distributions of input data remain consistent, the model's performance metrics (e.g., accuracy, precision) have significantly deteriorated. Which type of data or concept drift is most likely occurring in this scenario?
- A
Covariate Drift
- B
Label Drift
- C
Concept Drift
- D
Feature Drift
Show answer and explanation
Correct answer: C
Explanation
The observed issue in this scenario is a decline in model performance despite consistent feature distributions. This indicates a shift in the relationship between the input features and the target variable, which is characteristic of Concept Drift. Identifying and addressing Concept Drift is critical to maintaining the reliability of deployed machine learning models.
- A. Incorrect.
Covariate Drift refers to changes in the distribution of the input features (independent variables). Since the feature distributions remain consistent in this case, covariate drift is unlikely.
- B. Incorrect.
Label Drift refers to changes in the distribution of the target variable (dependent variable). There is no indication in the scenario of changes to the labels, so this is not the correct answer.
- C. Correct.
Concept Drift refers to changes in the relationship between the input features and the target variable, which directly impacts the model's performance. This aligns with the described issue where features remain consistent, but performance deteriorates.
- D. Incorrect.
Feature Drift is not a standard term in drift types for machine learning. The scenario explicitly states that the feature distributions remain consistent, so this is not relevant.