Google Professional Machine Learning Engineer Question 491
Select 3Google Cloud PlatformYou are deploying a machine learning model using Vertex AI to predict customer churn for a subscription-based service. The model's performance is critical and needs to be monitored for data drift and prediction quality over time. Which steps should you take to establish continuous evaluation of the model using Vertex AI Model Monitoring?
- A
Enable model monitoring and configure alerting thresholds for data drift and feature distribution changes.
- B
Set up a batch prediction job to manually evaluate the model's predictions every week.
- C
Specify the features to monitor and their expected statistical properties, such as feature skew and drift thresholds.
- D
Integrate Vertex AI Model Monitoring with a BigQuery table containing ground truth labels for prediction quality evaluation.
- E
Rely solely on the training dataset distribution to evaluate the model's performance over time.
Show answer and explanation
Correct answers: A, C, D
Explanation
To establish continuous evaluation metrics with Vertex AI Model Monitoring, you must enable model monitoring, configure alerting thresholds, and specify the features to monitor along with their statistical properties. Additionally, integrating ground truth labels in a BigQuery table allows you to assess prediction quality. These steps ensure that the deployed model is continuously monitored for data drift, feature skew, and overall performance degradation, which is critical for maintaining reliable predictions in production.
- A. Correct.
Correct: Enabling model monitoring and setting alerting thresholds are foundational steps in Vertex AI Model Monitoring for tracking metrics like data drift and feature distribution changes.
- B. Incorrect.
Incorrect: While batch prediction jobs can be useful, they are not a substitute for continuous monitoring. Continuous evaluation requires automated monitoring and alerting mechanisms.
- C. Correct.
Correct: Specifying features and their statistical properties helps Vertex AI Model Monitoring detect issues like feature skew and drift, which are critical for maintaining model performance.
- D. Correct.
Correct: Integrating ground truth labels in a BigQuery table allows Vertex AI Model Monitoring to assess prediction quality and verify model accuracy over time.
- E. Incorrect.
Incorrect: Relying solely on the training dataset distribution ignores real-world changes in the data and can lead to undetected model degradation.