Google Professional Machine Learning Engineer Question 21
Select 3Google Cloud PlatformYou are building a machine learning model to predict customer churn using BigQuery ML. Your dataset contains numerical, categorical, and text features. You want to ensure that your model performs optimally by addressing feature selection and preprocessing. Which steps should you take while using BigQuery ML for feature engineering or selection?
- A
Standardize numerical features by using the TRANSFORM clause in the CREATE MODEL statement.
- B
Use the ML.FEATURE_IMPORTANCE function after model training to identify important features.
- C
Manually encode categorical variables into one-hot representations before loading the data into BigQuery.
- D
Use the ML.PREDICT function to compute feature correlations prior to training the model.
- E
Leverage BigQuery ML's automatic preprocessing for supported input data types, such as category or numerical features.
Show answer and explanation
Correct answers: A, B, E
Explanation
BigQuery ML provides several built-in tools and functionalities for feature engineering and selection. The TRANSFORM clause allows for preprocessing transformations, such as standardization. Once the model is trained, ML.FEATURE_IMPORTANCE can help identify the most influential features. Additionally, BigQuery ML automatically preprocesses common feature types, like categorical and numerical data, reducing the need for manual effort. However, manual one-hot encoding is unnecessary, and ML.PREDICT is not intended for feature engineering.
- A. Correct.
This is correct. BigQuery ML supports standardizing numerical features using the TRANSFORM clause, which can improve model performance by ensuring that numerical features are on a similar scale.
- B. Correct.
This is correct. The ML.FEATURE_IMPORTANCE function in BigQuery ML can be used after training to identify which features have the most influence on the model's predictions, aiding in feature selection.
- C. Incorrect.
This is incorrect. BigQuery ML automatically handles categorical encoding for supported types, so manual one-hot encoding is unnecessary and could lead to redundant or incorrect preprocessing.
- D. Incorrect.
This is incorrect. The ML.PREDICT function is used for generating predictions, not for computing feature correlations or performing feature selection.
- E. Correct.
This is correct. BigQuery ML automatically preprocesses categorical and numerical features, reducing the need for manual feature engineering in many cases.