Google Professional Machine Learning Engineer Question 60
Select 3Google Cloud PlatformYou are working on a machine learning project to predict customer churn using Google Cloud AutoML Tables. Your dataset contains features like customer demographics, account activity logs, and support ticket history. Before importing the dataset into AutoML Tables, what steps should you prioritize to prepare the data for optimal model training?
- A
Ensure all target labels are consistent and correctly formatted.
- B
Remove any features with high correlation to the target variable to avoid data leakage.
- C
Fill missing values and ensure all features are in a machine-readable format.
- D
Eliminate all categorical features since AutoML Tables only supports numerical features.
- E
Balance the dataset if the target variable is highly imbalanced.
Show answer and explanation
Correct answers: A, C, E
Explanation
Preparing data for AutoML Tables involves ensuring the target labels are clean, handling missing or improperly formatted data, and addressing class imbalance when necessary. AutoML Tables automates many data preprocessing tasks, but proper preparation ensures the best model performance and avoids issues during training. Removing categorical features or features with high correlation to the target is unnecessary because AutoML Tables handles these cases during training.
- A. Correct.
Ensuring target labels are consistent and correctly formatted is critical because AutoML Tables requires a clean and well-defined target column to train the model effectively.
- B. Incorrect.
Removing features with high correlation to the target variable is not necessary for AutoML Tables as the service automatically handles feature selection and engineering. Removing such features could result in a loss of useful signal.
- C. Correct.
Filling missing values and ensuring features are in a machine-readable format is essential because AutoML Tables needs clean data to process effectively. Missing or improperly formatted data can lead to errors or suboptimal models.
- D. Incorrect.
Eliminating all categorical features is incorrect because AutoML Tables supports categorical features out of the box and can handle them automatically during preprocessing.
- E. Correct.
Balancing the dataset is important if the target variable is highly imbalanced. AutoML Tables provides tools to address imbalance, but preparing a balanced dataset before training often improves model performance.