Question 760
At an enterprise inventorying information for a new model, an AI project leader faces this situation: A delivery ETA model performs well downtown but has little training history for remote rural routes. Available records are not automatically relevant or authorized for training. Which plan change would best correct the situation?
Explanation
Correct Answer
In an enterprise inventorying information for a new model, the most direct response is: Compare training and planned deployment distributions across route types and conditions. This addresses: Dataset distribution must represent target deployment conditions to support generalization. Its acceptance evidence is: A segment-level analysis highlights undersampled regions and associated model performance risks. The project should also remember: Do not fabricate rare-route examples as if they were independent real observations.
Incorrect Answers
Add event lag checks, schema contracts and owned alerting to the ingestion pipeline. The appropriate correction for the reported case is: Compare training and planned deployment distributions across route types and conditions.
Review licenses and obtain explicit permitted-use approval before adding the source to training. This alternative is not the failing dependency described in the question.
Create a labeled comparison chart showing the agreed subgroup data or error distribution. Its verification is distinct from the reported condition: A delivery ETA model performs well downtown but has little training history for remote rural routes. Available records are not automatically relevant or authorized for training.