Databricks Machine Learning Associate Question 140
Single answerYou are analyzing a dataset in Databricks to prepare it for a machine learning model. During exploratory data analysis (EDA), you notice that the target variable has a significant number of missing values. What is the most appropriate action to take before proceeding with model training?
- A
Remove all rows where the target variable is missing
- B
Replace missing target values with the median of the target variable
- C
Use a separate imputation model to predict the missing target values
- D
Drop the target variable entirely and proceed without it
Show answer and explanation
Correct answer: A
Explanation
In supervised machine learning workflows, the target variable must be complete and accurate for effective training. Rows with missing target values should typically be removed to maintain the integrity of the dataset. Other approaches, like imputing the target variable, can introduce bias or lead to incorrect model behavior.
- A. Correct.
Removing rows with missing target values is generally the most appropriate action because machine learning models cannot be trained on rows where the target variable is missing. This ensures data consistency for supervised learning models.
- B. Incorrect.
Replacing missing target values with the median can introduce bias and is not recommended, as it distorts the relationship between the features and the target variable.
- C. Incorrect.
Using a separate imputation model for the target variable is not a standard practice in supervised learning workflows, as it could lead to unreliable target values and model performance issues.
- D. Incorrect.
Dropping the target variable entirely would make it impossible to train a supervised learning model, as the target variable is essential for such tasks.