Exam Professional Machine Learning Engineer topic 1 question 291 discussion - ExamTopics

Problem:

A hospital deployed a Vertex AI model predicting patient risk. Concerns exist about changing patient demographics affecting feature interactions and prediction accuracy. The goal is cost-effective monitoring for feature interaction changes and feature importance understanding.

Options:

• A: Feature drift monitoring with a sampling rate of 1 and weekly frequency.
• B: Feature drift monitoring with a sampling rate of 0.1 and weekly frequency.
• C: Feature attribution drift monitoring with a sampling rate of 1 and weekly frequency.
• D: Feature attribution drift monitoring with a sampling rate of 0.1 and weekly frequency.

Solution:

The suggested answer is D: Creating a feature attribution drift monitoring job with a sampling rate of 0.1 and weekly frequency. This option balances cost-effectiveness (sampling rate of 0.1) with the need to monitor both feature drift and attribution, addressing the concern of changing feature interactions.