MLA-C01 Question 361
Select 3You have deployed a machine learning model to an Amazon SageMaker endpoint. Over time, you notice that model predictions are becoming less accurate due to changes in the underlying data distribution. What actions should you take to ensure the model remains accurate and reliable?
- A
Enable Amazon SageMaker Model Monitor to track data drift and set up alerts for significant deviations.
- B
Retrain the model periodically using new data to capture changes in the data distribution.
- C
Configure Amazon CloudWatch Logs to automatically retrain the model without manual intervention when data drift is detected.
- D
Manually inspect the training and inference datasets to identify differences and adjust preprocessing steps.
- E
Use SageMaker Clarify to detect and mitigate bias in the data before retraining the model.
Show answer and explanation
Correct answers: A, B, D
Explanation
To maintain the accuracy and reliability of a deployed machine learning model, it is important to monitor for data drift using tools like SageMaker Model Monitor, retrain the model with updated datasets to adapt to changing data distributions, and ensure data preprocessing steps remain consistent. While SageMaker Clarify is useful for detecting and mitigating bias, it is not directly relevant to addressing data drift in this scenario. Automation of retraining requires custom workflows and is not automatically handled by CloudWatch Logs.
- A. Correct.
Correct. Amazon SageMaker Model Monitor can continuously track data drift and help you set up alerts for deviations, allowing you to take corrective actions promptly.
- B. Correct.
Correct. Retraining the model with updated datasets ensures it adapts to changes in the data distribution and maintains high accuracy.
- C. Incorrect.
Incorrect. While Amazon CloudWatch Logs can be used to monitor logs, it cannot directly retrain the model. Retraining requires manual setup or automation through custom workflows.
- D. Correct.
Correct. Manually inspecting training and inference datasets can help identify specific issues, such as feature mismatches or data preprocessing problems, which can improve the model's accuracy.
- E. Incorrect.
Incorrect. SageMaker Clarify is primarily used for bias detection and explainability, not for detecting or addressing data drift directly.