Exam Associate Data Practitioner topic 1 question 56 discussion

The best option is C. Evaluate data drift. Option C is best because data drift specifically refers to changes in data distribution over time, directly indicating if the current serving data differs from the historical data the model was trained on, thus signaling a need for retraining. Option A (Compare models) is incorrect because comparing models doesn't directly assess changes in serving data. Option B (Evaluate skewness) is incorrect because skewness is a static data property, not a comparison of data over time. Option D (Compare confusion matrix) is incorrect because it evaluates model performance change, an indirect consequence of data drift, not the drift itself. Therefore, Option C is the most direct method to determine if data changes necessitate model retraining.