Google Professional Machine Learning Engineer Question 25
Single answerGoogle Cloud PlatformYou are working on a regression model using BigQuery ML to predict house prices based on various features such as square footage, number of bedrooms, and location. During the feature engineering process, you notice that some features, such as 'property_type' and 'zip_code', are categorical. How can you handle these categorical features effectively in BigQuery ML to ensure the model can utilize them?
- A
Use the TRANSFORM clause in BigQuery ML to apply one-hot encoding directly in the SQL query.
- B
Manually convert the categorical features into numerical values using external preprocessing tools before importing the data into BigQuery.
- C
Use the TRANSFORM clause in BigQuery ML to apply bucketing to group similar categories together.
- D
Allow BigQuery ML to automatically handle categorical features without any additional preprocessing.
Show answer and explanation
Correct answer: A
Explanation
In BigQuery ML, the TRANSFORM clause is a powerful tool for feature engineering that allows users to apply transformations directly in SQL queries. For categorical features, one-hot encoding is a common approach to convert them into numerical features, and this can be accomplished efficiently using the TRANSFORM clause. This ensures that the model can interpret and utilize the categorical data effectively without requiring external preprocessing.
- A. Correct.
Correct: BigQuery ML provides a convenient TRANSFORM clause that allows you to apply transformations like one-hot encoding directly in SQL, which is ideal for handling categorical features in your data.
- B. Incorrect.
Incorrect: While external preprocessing tools can be used, BigQuery ML already provides built-in capabilities for feature engineering, making external tools unnecessary in this case.
- C. Incorrect.
Incorrect: Bucketing is used to group numerical values into ranges or bins. It is not the appropriate technique for handling categorical variables.
- D. Incorrect.
Incorrect: Although BigQuery ML can automatically handle some feature engineering, relying solely on automatic handling may lead to suboptimal results, especially for categorical features that require specific transformations like one-hot encoding.