PRODUCTION SCENARIO
An existing Model Monitor customer sees no data quality violations on a production endpoint, but business users report a rising number of incorrect predictions. Confirmed outcome labels for each prediction land in Amazon S3 about seven days after the inference.
Which Model Monitor schedule will detect the degradation?
Answering here is anonymous. Nothing is saved unless you sign in.
Show answer and explanation
Answer: A model quality schedule that merges predictions with the ground truth labels
Model quality monitoring merges captured predictions with the ground truth labels in Amazon S3 and scores accuracy against a baseline, the only one of the four types that observes whether predictions were right. The seven-day delay changes when the job runs, not whether it works. Data quality watches inputs, bias drift watches subgroup disparity, and attribution drift watches feature weight.