A binary classification dataset has 95% negative class and 5% positive class samples. What is the most direct consequence of training a model on this imbalanced data without any mitigation?

The model will produce unbiased predictions because the training algorithm corrects for class frequency.

The model will overfit the minority class because it has fewer samples.

The model will be biased toward the majority class and may achieve high accuracy while failing to identify minority class instances.

The model will automatically adjust its decision threshold to favor the minority class.

A data scientist is training a gradient boosting model for a rare event prediction. They want to use class weights to penalize misclassifications of the minority class. Which of the following is the most appropriate way to implement this in a typical ML framework?

Set the learning rate to a very low value

Pass a class_weight parameter to the model constructor

Use a custom loss function that ignores majority class errors

Manually duplicate minority class samples in the training set

What is a primary disadvantage of random undersampling the majority class to address data imbalance?

It increases the risk of overfitting due to duplication.

It creates synthetic samples that may not represent real data.

It requires more computational resources than oversampling.

It can lead to loss of potentially valuable information from the majority class.

Why is the F1-score often preferred over accuracy for imbalanced classification problems?

F1-score does not require a threshold, unlike accuracy.

F1-score is always higher than accuracy, making it easier to interpret.

F1-score is the harmonic mean of precision and recall, providing a balanced measure that is sensitive to both false positives and false negatives.

F1-score is only used for multi-class problems, not binary.

DatabricksCertified Machine Learning Associate

Domain 3Objective 2

Data Preparation and Imbalance MACHINE-LEARNING-ASSOCIATE Practice Questions (Page 4)

Part of the Section 3: Model Development domain, which makes up ~23% of our current practice bank. Databricks does not publish an official question count, but from its 90-minute exam (~35–60 total, ~8–14 in this domain), expect 2–4 from this objective — we provide 21 practice questions to prepare you well beyond it. (estimate)

21questions here

5free pages

4concepts

Questions 16–20

16foundation · easy
A binary classification dataset has 95% negative class and 5% positive class samples. What is the most direct consequence of training a model on this imbalanced data without any mitigation?
Select an answer first
17application · medium
A data scientist is training a gradient boosting model for a rare event prediction. They want to use class weights to penalize misclassifications of the minority class. Which of the following is the most appropriate way to implement this in a typical ML framework?
Select an answer first
18foundation · easy
What is a primary disadvantage of random undersampling the majority class to address data imbalance?
Select an answer first
19foundation · easy
Why is the F1-score often preferred over accuracy for imbalanced classification problems?
Select an answer first
20application · medium
A team has a dataset with 50,000 majority class samples and 500 minority class samples. They want to use random undersampling to balance the classes. How many majority class samples should they randomly select to achieve a 1:1 ratio?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to MACHINE-LEARNING-ASSOCIATE

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-ASSOCIATE” is a trademark of its owner, used for identification only.