A data engineer is using Spark ML to one-hot encode a categorical column 'size' with values 'S', 'M', 'L'. They use StringIndexer with default settings. What is the output of StringIndexer for a row with 'M'?

A numeric index, such as 1, representing the category 'M'.

A binary vector with a 1 at the position for 'M'.

A one-hot encoded vector of length 3.

How does one-hot encoding typically affect model interpretability compared to label encoding?

Label encoding is always more interpretable because it preserves the ordinal relationship.

Both methods have identical interpretability because they encode the same information.

One-hot encoding improves interpretability by giving each category its own coefficient or feature importance.

One-hot encoding reduces interpretability because it creates too many features to analyze.

How does Spark ML's StringIndexer handle unseen categories that appear in new data during inference?

It ignores unseen categories and leaves the row without a value.

It throws an error unless handleInvalid is set to 'skip' or 'keep'.

It automatically assigns a new index to each unseen category.

It maps all unseen categories to the most frequent category.

A data team is building a Spark ML pipeline with multiple stages: imputation, one-hot encoding, scaling, and model training. They want to ensure that the one-hot encoding stage is applied only to categorical columns and not to numeric columns. What is the best way to achieve this in a Pipeline?

Use a single StringIndexer and OneHotEncoder on the entire dataframe, which will automatically skip numeric columns.

Use a Pipeline with separate branches: one for categorical columns (StringIndexer + OneHotEncoder) and one for numeric columns, then combine with VectorAssembler.

Apply OneHotEncoder to all columns, including numeric ones, and let the model handle the extra features.

Use a custom UDF to one-hot encode all columns in a single step, avoiding the need for separate stages.

DatabricksCertified Machine Learning Associate

Domain 2Objective 7

Use One-Hot Encoding for Categorical Features MACHINE-LEARNING-ASSOCIATE Practice Questions (Page 4)

Part of the Section 2: Data Processing domain, which makes up ~27% of our current practice bank. Databricks does not publish an official question count, but from its 90-minute exam (~35–60 total, ~9–16 in this domain), expect 1–2 from this objective — we provide 23 practice questions to prepare you well beyond it. (estimate)

23questions here

5free pages

5concepts

Questions 16–20

16application · easy
A data engineer is using Spark ML to one-hot encode a categorical column 'size' with values 'S', 'M', 'L'. They use StringIndexer with default settings. What is the output of StringIndexer for a row with 'M'?
Select an answer first
17foundation · easy
How does one-hot encoding typically affect model interpretability compared to label encoding?
Select an answer first
18foundation · easy
How does Spark ML's StringIndexer handle unseen categories that appear in new data during inference?
Select an answer first
19expert · hard
A data team is building a Spark ML pipeline with multiple stages: imputation, one-hot encoding, scaling, and model training. They want to ensure that the one-hot encoding stage is applied only to categorical columns and not to numeric columns. What is the best way to achieve this in a Pipeline?
Select an answer first
20foundation · easy
In Spark ML, which transformer is used to convert a column of categorical indices into a vector of binary indicators?
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.