Data drift vs concept drift?

Data drift vs concept drift?

• A. Data drift is when input data distributions change over time; concept drift is when the relationship between inputs and the target changes.
• B. Data drift is when the target changes; concept drift is when inputs change.
• C. Data drift and concept drift describe the same phenomenon.
• D. Data drift refers to model performance decay due to aging data.

Answer: (A)

Data drift vs concept drift centers on what changes in the data you’re using. Data drift means the input feature distributions change over time (the distribution P(X) shifts) while the relationship between inputs and the target stays the same (P(Y|X) remains stable). Concept drift means the relationship itself changes (P(Y|X) changes), so the same inputs can lead to different outputs over time. For example, if the types of features you see in transactions evolve but the way those features indicate fraud stays the same, that’s data drift. If the fraud indicators themselves stop signaling fraud in the same way, that’s concept drift. The other statements mix up what changes or are too vague about performance, so they don’t capture the precise distinction.

Data drift vs concept drift centers on what changes in the data you’re using. Data drift means the input feature distributions change over time (the distribution P(X) shifts) while the relationship between inputs and the target stays the same (P(Y|X) remains stable). Concept drift means the relationship itself changes (P(Y|X) changes), so the same inputs can lead to different outputs over time. For example, if the types of features you see in transactions evolve but the way those features indicate fraud stays the same, that’s data drift. If the fraud indicators themselves stop signaling fraud in the same way, that’s concept drift. The other statements mix up what changes or are too vague about performance, so they don’t capture the precise distinction.