PMLE Monitoring ML Solutions Practice Question

A company has deployed a model to Vertex AI Endpoints and wants to monitor for feature drift using Jensen-Shannon divergence. They have set a threshold of 0.1. After one week, the monitoring job reports a divergence of 0.15 for a feature. What should the engineer do next to diagnose which features are contributing to the drift?

Answer choices

Why each option matters

Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.

Correct answer & explanation

✓

Use Vertex AI Explainability to compute feature attributions and identify drifted features

To identify which features are drifting, the engineer can use Vertex AI Explainability to compute feature attributions (e.g., SHAP values) and correlate them with drift metrics.

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

✗
Deploy a new model version immediately
Why it's wrong here
Deploying a new model may not address the root cause if drift is due to data changes.
✓
Use Vertex AI Explainability to compute feature attributions and identify drifted features
Why this is correct
Correct: Explainability provides feature importance, helping to pinpoint which features are driving drift.
✗
Check the model's confusion matrix in BigQuery
Why it's wrong here
Confusion matrix is for model quality monitoring, not feature drift analysis.
✗
Increase the sampling rate to capture more data
Why it's wrong here
Sampling rate affects detection sensitivity but does not identify which features are responsible.

Go deeper

PMLE Monitoring ML Solutions Practice Question

Why each option matters

Option-by-option breakdown

Related to this question