Use One-Hot Encoding for Categorical Features MACHINE-LEARNING-ASSOCIATE Practice Questions (Page 1)

Questions 1–5

1. 1expert · hard

   A data scientist is building a churn prediction model with a logistic regression classifier. The dataset has a categorical feature 'product_category' with 500 categories. They are concerned about model interpretability and performance. Which approach best balances these concerns?

   • AUse target encoding, replacing each category with the mean target value, which reduces dimensionality and can improve performance.
   • BUse one-hot encoding with all 500 categories, accepting the high dimensionality for the sake of interpretability.
   • CGroup rare categories into an 'Other' category before one-hot encoding, reducing the number of columns while retaining interpretability for major categories.
   • DUse feature hashing to reduce the dimensionality to a fixed number of features, trading off interpretability for performance.
   Reveal Solution Discussion Select an answer first
2. 2expert · hard

   A data team is building a Spark ML pipeline for a classification task. They have a categorical column 'region' with 10 categories. They are considering whether to use one-hot encoding or leave the column as a string and rely on the model to handle it. Which statement is correct?

   • ATree-based models in Spark ML can handle string columns natively without encoding.
   • BLeaving the column as a string is acceptable because Spark ML automatically converts strings to numeric indices.
   • CMost Spark ML models, including linear models, cannot handle string columns directly, so one-hot encoding is necessary.
   • DOne-hot encoding is only needed for linear models, not for tree-based models.
   Reveal Solution Discussion Select an answer first
4. 3foundation · easy

   Which type of column is most appropriate for one-hot encoding?

   • AA continuous numeric column such as age or income
   • BA categorical column with a small number of distinct values
   • CA column with missing values that need imputation
   • DA column with high cardinality, such as user IDs
   Reveal Solution Discussion Select an answer first
5. 4foundation · easy

   In a Spark ML pipeline, which stage typically comes immediately before OneHotEncoder?

   • AStandardScaler
   • BStringIndexer
   • CVectorAssembler
   • DRandomForestClassifier
   Reveal Solution Discussion Select an answer first
6. 5foundation · easy

   Which statement accurately describes the output of one-hot encoding for a categorical column with three distinct categories?

   • AThree binary columns, each indicating the presence of one category.
   • BA single column with values 0, 0.5, and 1 representing the categories.
   • CTwo binary columns, using a combination of bits to represent the three categories.
   • DA single column with values 1, 2, and 3 representing the categories.
   Reveal Solution Discussion Select an answer first