Google Professional Machine Learning Engineer Question 5
Select 4Google Cloud PlatformYou are a data scientist at a retail company and want to build a machine learning model to predict customer churn using BigQuery ML. You have a dataset stored in BigQuery with millions of rows and several features. Which steps should you perform to ensure the model is developed effectively while leveraging BigQuery ML's capabilities?
- A
Use the CREATE MODEL statement in BigQuery ML to define the model type and target column.
- B
Preprocess the dataset by writing SQL queries to handle missing values and normalize features within BigQuery.
- C
Export the dataset from BigQuery to a third-party tool for feature engineering and model training.
- D
Leverage built-in feature importance tools in BigQuery ML to evaluate which features contribute the most to the model.
- E
Use a separate validation dataset within BigQuery to evaluate model performance after training.
Show answer and explanation
Correct answers: A, B, D, E
Explanation
BigQuery ML enables end-to-end machine learning workflows within BigQuery, making it unnecessary to export data for external preprocessing or training. You can preprocess the data using SQL, train models using the CREATE MODEL statement, evaluate feature importance, and validate performance, all within the BigQuery environment.
- A. Correct.
Correct: The CREATE MODEL statement is the core feature of BigQuery ML, allowing you to specify model type (e.g., logistic regression) and the target column for prediction.
- B. Correct.
Correct: BigQuery ML supports SQL-based preprocessing, enabling you to handle missing values, normalize data, or create new features directly in BigQuery.
- C. Incorrect.
Incorrect: Exporting the dataset to a third-party tool contradicts the goal of leveraging BigQuery ML's native capabilities for model development and training.
- D. Correct.
Correct: BigQuery ML provides built-in feature importance tools for specific model types, helping you understand which features are most relevant to the model.
- E. Correct.
Correct: Using a separate validation dataset within BigQuery ensures proper model evaluation and avoids overfitting to the training data.