Question

A retail company runs a SageMaker real-time endpoint for a credit approval model. Over recent weeks, the overall accuracy on ground-truth data remains stable, but the business team suspects the model is relying on different input features than during training, which could signal changing customer behavior. The ML engineer wants an automated way to detect when the relative importance of input features shifts significantly compared to the training baseline, and to alert the team without manually recomputing feature importance. Which approach best meets this requirement?

Accepted Answer

Enable SageMaker Model Monitor feature attribution drift monitoring, which uses SageMaker Clarify (SHAP) to compare live feature importance against the baseline and emits CloudWatch metrics for alarms.

Answer

Enable SageMaker Model Monitor data quality monitoring to compare feature distributions against the baseline and trigger a CloudWatch alarm on statistical drift.

Answer

Enable SageMaker Model Monitor model quality monitoring with ground-truth labels to track accuracy degradation and alarm when metrics fall below a threshold.

Answer

Schedule a periodic SageMaker Processing job that retrains the model and compares the new coefficients to the previous version, alerting if they differ.