Databricks Machine Learning Associate Question 173
Select 2When preparing a dataset for machine learning, why is it important to add indicator variables for missing values that have been imputed or replaced?
- A
Indicator variables can help the model capture potential patterns in missingness that may be predictive.
- B
Adding indicator variables ensures the imputed values are treated as genuine observations by the model.
- C
It prevents the machine learning model from overfitting to the imputed values.
- D
Missing values might carry important information about the underlying data that indicator variables can encode.
- E
Indicator variables are required to identify and remove missing values from the dataset entirely.
Show answer and explanation
Correct answers: A, D
Explanation
Adding indicator variables for missing values that have been imputed or replaced is important because missingness can itself be an informative signal for the model. For example, the fact that a value is missing might correlate with the target variable or other features. Indicator variables allow the model to leverage this potential relationship, improving predictive performance. However, indicator variables are not used to treat imputed values as genuine observations or to remove missing values entirely.
- A. Correct.
Indicator variables can help the model capture potential patterns in missingness that may be predictive, such as missingness itself being correlated with the target variable.
- B. Incorrect.
This is incorrect. Adding indicator variables does not ensure the imputed values are treated as genuine observations; instead, it provides additional context to the model on where missing values existed.
- C. Incorrect.
This is incorrect. Indicator variables do not directly prevent overfitting; their purpose is to encode information about missingness, not regularize the model.
- D. Correct.
Missing values might carry important information, such as systematic reasons for why data is missing, and indicator variables help encode this information for the model.
- E. Incorrect.
This is incorrect. Indicator variables are not used to remove missing values but to provide additional information about where missingness occurred.