nerdexam
Google

PROFESSIONAL-MACHINE-LEARNING-ENGINEER · Question #302

You work for a hospital. You received approval to collect the necessary patient data, and you trained a Vertex AI tabular AutoML model that calculates patients' risk score for hospital admission…

The correct answer is D. Create a feature attribution drift monitoring job. Set the sampling rate to 0.1 and the monitoring. To monitor changes in feature interactions and importance while minimizing cost for a deployed model, create a feature attribution drift monitoring job with a low sampling rate and daily frequency.

Submitted by fatema_kw· Apr 18, 2026Monitoring, optimizing, and maintaining ML solutions

Question

You work for a hospital. You received approval to collect the necessary patient data, and you trained a Vertex AI tabular AutoML model that calculates patients' risk score for hospital admission. You deployed the model. However, you're concerned that patient demographics might change over time and alter the feature interactions and impact prediction accuracy. You want to be alerted if feature interactions change, and you want to understand the importance of the features for the predictions. You want your alerting approach to minimize cost. What should you do?

Options

  • ACreate a feature drift monitoring job. Set the sampling rate to 1 and the monitoring frequency to
  • BCreate a feature drift monitoring job. Set the sampling rate to 0.1 and the monitoring frequency to
  • CCreate a feature attribution drift monitoring job. Set the sampling rate to 1 and the monitoring
  • DCreate a feature attribution drift monitoring job. Set the sampling rate to 0.1 and the monitoring

How the community answered

(38 responses)
  • A
    8% (3)
  • B
    3% (1)
  • C
    13% (5)
  • D
    76% (29)

Why each option

To monitor changes in feature interactions and importance while minimizing cost for a deployed model, create a feature attribution drift monitoring job with a low sampling rate and daily frequency.

ACreate a feature drift monitoring job. Set the sampling rate to 1 and the monitoring frequency to

Feature drift monitoring only checks for changes in individual feature distributions, not their interactions or importance to predictions, and a high sampling rate/frequency increases cost.

BCreate a feature drift monitoring job. Set the sampling rate to 0.1 and the monitoring frequency to

Feature drift monitoring does not address the concern of changing 'feature interactions' or 'importance of features for predictions', which requires attribution analysis.

CCreate a feature attribution drift monitoring job. Set the sampling rate to 1 and the monitoring

While feature attribution drift monitoring is correct for the problem, a sampling rate of 1 and monitoring frequency of 1 hour would incur significantly higher costs than necessary for this use case.

DCreate a feature attribution drift monitoring job. Set the sampling rate to 0.1 and the monitoringCorrect

Feature attribution drift monitoring specifically detects changes in how much each feature contributes to predictions, which indicates altered feature interactions and importance. Setting a sampling rate of 0.1 and a monitoring frequency of 1 day significantly minimizes compute costs compared to higher rates or frequencies, aligning with the cost minimization goal.

Concept tested: Vertex AI Model Monitoring feature attribution drift

Source: https://cloud.google.com/vertex-ai/docs/model-monitoring/overview#drift-detection

Topics

#ML Model Monitoring#Feature Attribution Drift#Cost Optimization#Vertex AI

Community Discussion

No community discussion yet for this question.

Sign up to join the discussion
Full PROFESSIONAL-MACHINE-LEARNING-ENGINEER Practice