An ML engineer is setting up a monitor on a time series table. They need to add a custom metric that calculates the ratio of 'returns' to 'sales' for each day. What must they specify when defining this custom metric?

Only the SQL expression and the baseline period

Metric name, type, and SQL expression for the computation

Only the metric name and the alert threshold

The metric name and the notification channel

A credit-risk model uses a numerical feature 'loan_amount' and a categorical feature 'employment_type'. The ML engineer creates a Lakehouse Monitor on the inference table and reviews the drift metrics table. The KS test for 'loan_amount' reports a p-value of 0.001, and the chi-squared test for 'employment_type' reports a p-value of 0.6. The team's threshold for alerting is p < 0.05. Which conclusion is correct?

Both features show statistically significant drift and should be investigated.

The KS test is invalid for numerical data, so the results cannot be interpreted.

Neither feature shows drift because the p-values are not below 0.01.

Only 'loan_amount' shows statistically significant drift; 'employment_type' does not.

A team monitors a recommendation model. The overall accuracy is stable at 85%, but when they slice the data by user segment, they see that accuracy for 'new_users' has dropped from 80% to 60% over the past month. What should they do?

Investigate the 'new_users' segment for specific issues, such as data drift or model bias.

Retrain the model on all data to improve overall accuracy.

Ignore the segment drop because overall accuracy is stable.

Change the alert threshold for accuracy to 0.5.

In Lakehouse Monitoring, what is the primary purpose of setting an alert rule on a drift metric?

To notify stakeholders when the metric exceeds a predefined threshold

To automatically retrain the model when drift is detected

To log additional data when drift is detected

To change the baseline of the monitor automatically

DatabricksCertified Machine Learning Professional

Domain 2Objective 5

Drift Detection and Lakehouse Monitoring MACHINE-LEARNING-PROFESSIONAL Practice Questions (Page 5)

Part of the ML Ops domain, which accounts for 44% of the MACHINE-LEARNING-PROFESSIONAL exam. Databricks does not publish an official question count, but from its 120-minute exam (~50–80 total, ~22–35 in this domain), expect 4–7 from this objective — we provide 34 practice questions to prepare you well beyond it. (estimate)

34questions here

7free pages

10concepts

44%of the exam

Questions 21–25

21application · medium
An ML engineer is setting up a monitor on a time series table. They need to add a custom metric that calculates the ratio of 'returns' to 'sales' for each day. What must they specify when defining this custom metric?
Select an answer first
22foundation · easy
An increase in which metric would most directly indicate that a model endpoint is returning more errors?
Select an answer first
23application · medium
A credit-risk model uses a numerical feature 'loan_amount' and a categorical feature 'employment_type'. The ML engineer creates a Lakehouse Monitor on the inference table and reviews the drift metrics table. The KS test for 'loan_amount' reports a p-value of 0.001, and the chi-squared test for 'employment_type' reports a p-value of 0.6. The team's threshold for alerting is p < 0.05. Which conclusion is correct?
Select an answer first
24expert · hard
A team monitors a recommendation model. The overall accuracy is stable at 85%, but when they slice the data by user segment, they see that accuracy for 'new_users' has dropped from 80% to 60% over the past month. What should they do?
Select an answer first
25foundation · easy
In Lakehouse Monitoring, what is the primary purpose of setting an alert rule on a drift metric?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to MACHINE-LEARNING-PROFESSIONAL

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by Databricks. “MACHINE-LEARNING-PROFESSIONAL” is a trademark of its owner, used for identification only.

Drift Detection and Lakehouse Monitoring MACHINE-LEARNING-PROFESSIONAL Practice Questions (Page 5)

Questions 21–25

An ML engineer is setting up a monitor on a time series table. They need to add a custom metric that calculates the ratio of 'returns' to 'sales' for each day. What must they specify when defining this custom metric?

An increase in which metric would most directly indicate that a model endpoint is returning more errors?

A team monitors a recommendation model. The overall accuracy is stable at 85%, but when they slice the data by user segment, they see that accuracy for 'new_users' has dropped from 80% to 60% over the past month. What should they do?

In Lakehouse Monitoring, what is the primary purpose of setting an alert rule on a drift metric?

Keep studying — ML Ops