Engineer Features for Models SNOWPRO-ADVANCED-DATA-SCIENTIST Practice Questions (Page 3)

Questions 11–15

1. 11expert · hard

   A data scientist is building a production pipeline for a linear regression model that predicts house prices. The training data has 5% missing values in 'square_footage' and 40% missing values in 'year_renovated'. The test data arrives in streaming batches. The scientist must choose an imputation strategy that minimizes bias and is robust to distribution shift. Which approach is best?

   • AImpute both columns with the mean of their respective training distributions.
   • BUse a k-NN imputer on both columns using all other features.
   • CImpute 'square_footage' with the training median and 'year_renovated' with a constant '0' flag indicating no renovation.
   • DDrop all rows containing any missing value before training.
   Reveal Solution Discussion Select an answer first
2. 12foundation · easy

   Which feature selection method evaluates features independently of any machine learning model?

   • APrincipal component analysis (PCA)
   • BFilter method
   • CWrapper method
   • DEmbedded method
   Reveal Solution Discussion Select an answer first
4. 13application · medium

   A data scientist is working on a tabular dataset with 50 columns and limited time to manually engineer features. The scientist wants to systematically generate a large set of candidate features and then let a model select the useful ones. Which approach best fits this workflow?

   • AManually create a few interaction features based on domain knowledge.
   • BUse a random forest to select the top 10 original features and discard the rest.
   • CApply PCA to the raw columns and use the principal components as features.
   • DUse the Featuretools library to automatically generate deep feature sets from the relational data.
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   A data scientist is building a logistic regression model to predict customer churn. The dataset contains a 'tenure_months' column ranging from 1 to 72, and an 'age' column ranging from 18 to 90. The model is converging slowly and the coefficient magnitudes are difficult to interpret. Which action should the data scientist take to address both issues?

   • AApply Min-Max scaling to both 'tenure_months' and 'age' so they fall in the [0, 1] range.
   • BUse a tree-based model instead, since it does not require feature scaling.
   • CApply log transformation to 'tenure_months' and Min-Max scaling to 'age'.
   • DApply StandardScaler to 'tenure_months' only, leaving 'age' unchanged.
   Reveal Solution Discussion Select an answer first
6. 15foundation · easy

   Which transformation technique maps numeric features to a fixed range, typically [0, 1]?

   • AStandardization (z-score)
   • BMin-max scaling
   • CLog transformation
   • DOne-hot encoding
   Reveal Solution Discussion Select an answer first