A company deployed a machine learning (ML) model on the company website to predict real estate prices. Several months after deployment, an ML engineer notices that the accuracy of the model has…

A gradual decrease in model accuracy over time indicates model drift, which can be mitigated by incremental training with new, fresh data to update the model. Amazon SageMaker Model Monitor is designed to continuously monitor the performance of deployed models, detect data drift, model quality degradation, and bias, and can send notifications for any detected performance issues, directly addressing both requirements.

Amazon SageMaker Model Governance primarily focuses on managing model lifecycles, compliance, and auditing, not directly on detecting runtime performance degradation or automatically updating models in response to drift.

Amazon SageMaker Debugger is used during model training to analyze and debug training issues (e.g., vanishing gradients, overfitting) and does not monitor deployed models for gradual accuracy degradation or send notifications for runtime performance issues.

While incremental training is part of the solution, using only data from the previous several months might discard valuable historical patterns, and this choice does not address the crucial requirement of receiving notifications for future performance issues through a monitoring service.