HikeLetter | PMP Training & Certification Online Course | Project Management Learning

1

For a customer churn project, a data scientist needs to track parameters, metrics, and artifacts across several model runs. Which capability is most appropriate?

A. MLflow experiment tracking B. A SQL warehouse name C. A dashboard color setting D. A cluster tag alone

2

When building a repeatable ML experiment, a classification dataset has a rare positive class. Which evaluation measure is often more informative than accuracy alone?

A. Precision, recall, F1, or PR-AUC, chosen for the business objective B. The number of notebook cells C. The model artifact file size alone D. The alphabetical order of feature names

3

For a team preparing a deployment, a feature is calculated using information that would only be known after the prediction time. What risk does this create?

A. Target leakage B. Improved unbiased evaluation by definition C. Guaranteed model fairness D. A reduction in all missing values

4

For an ML application with changing data, a model's prediction quality degrades after deployment as the data distribution changes. What should the team implement?

A. Monitor input and performance signals and define retraining or investigation triggers B. Assume the model remains accurate forever C. Delete historical metrics D. Only monitor the endpoint's display name

5

For a production scoring workflow, a training dataset contains an identifier that has no predictive meaning and is unique for every row. What should the practitioner do first?

A. Assess whether the identifier leaks information or adds value, and usually exclude it from features B. Use it as the target label automatically C. Duplicate it to increase feature importance D. Sort the dataset by identifier and assume the model will generalize

6

In a model development review, a team wants repeatable model training with consistent preprocessing. What is a strong practice?

A. Package preprocessing and estimator steps into a reproducible pipeline B. Apply different manual transformations each run C. Edit the validation labels after predictions D. Store only screenshots of the results

7

In a multi-run experiment, a model registry is used to manage candidate and production model versions. Why is this useful?

A. It supports versioning, lineage, review, and controlled promotion B. It automatically guarantees every model is accurate C. It removes the need for validation D. It replaces data governance entirely

8

In a governed Databricks workspace, a team wants to compare feature importance across model versions. What should it ensure first?

A. Use compatible evaluation data, definitions, and interpretation methods B. Compare scores from unrelated datasets without context C. Rename the features randomly D. Use training accuracy as the only evidence

9

For a business forecasting use case, a model performs very well on training data but poorly on a held-out validation set. What is the most likely issue to investigate?

A. Overfitting and possible data leakage B. The dashboard theme C. The model's display name D. Whether the notebook title is descriptive

10

When evaluating a new feature set, two runs use different hyperparameters and the team wants to compare them. What should be logged?

A. Parameters, relevant metrics, artifacts, and dataset/code context B. Only the run's display color C. Only the time of day D. A free-text note with no run association

Databricks Certified Machine Learning Engineer Associate

Overview

Course Description

Course Questions