MLOps - Drifts!

Which type of drift affects the relationship between input and target variables?

• Data Drift
• Feature Drift
• Model Drift
• Concept Drift

Answer explanation

Concept drift refers to changes in the underlying relationship between input data and the target variable over time.

Which term is synonymous with data drift?

• Label shift
• Concept shift
• Covariate shift
• Output shift

Answer explanation

Covariate shift is another name for data drift, where the distribution of input features changes while the relationship to the target remains the same.

Which of the following is NOT a type of concept drift?

• Sudden
• Gradual
• Incremental
• Irregular

Answer explanation

Irregular is not a standard classification; the commonly recognized types include sudden, gradual, and incremental.

What is the key difference between concept drift and data drift?

• Concept drift relates to output; data drift to input
• Data drift happens slowly; concept drift happens instantly
• Concept drift is reversible; data drift is not
• There is no real difference

Answer explanation

Concept drift involves changes in the relationship between input and output, while data drift refers to changes in the input data distribution.

A change in the way spam emails are written over time can lead to:

• Sudden Data Drift
• Recurring Concept Drift
• Sudden Concept Drift
• Gradual Data Drift

Answer explanation

Sudden changes in spam patterns affect the target concept, qualifying as sudden concept drift.

What is one major consequence of model drift in healthcare?

• Longer training time
• Incorrect diagnoses
• More frequent retraining
• Reduced model complexity

Answer explanation

Model drift can lead to inaccurate predictions, such as misdiagnoses, which is critical in healthcare settings.

Is it possible to eliminate model drift completely?

• Yes, using deep learning
• No, only mitigate it
• Yes, with enough data
• Yes, using ensemble models

Answer explanation

Model drift cannot be completely eliminated, only mitigated through monitoring and updates.

What happens to model accuracy over time if drift is not managed?

• It improves
• It stays the same
• It fluctuates
• It degrades

Answer explanation

Without addressing drift, the model's relevance decreases over time, leading to degraded performance.

During an economic downturn, previously low-risk borrowers default more. What drift is this?

• Data Drift
• Concept Drift
• Feature Drift
• Model Drift

Answer explanation

The relationship between borrower features and loan outcomes changes, indicating concept drift.

Fraudsters change tactics, altering the patterns of fraud. What drift does this represent?

• Data Drift
• Concept Drift
• Model Drift
• No Drift

Answer explanation

The decision boundary evolves as new fraudulent behaviors emerge—this is concept drift.