Google Professional Machine Learning Engineer Question 64
Select 3Google Cloud PlatformYou are a Machine Learning Engineer tasked with preparing data for a retail sales prediction model using Google Cloud AutoML Tables. The dataset contains thousands of rows with missing values, categorical variables, and irrelevant columns like 'Customer Notes.' Which of the following steps should you take to ensure the data is properly prepared for AutoML Tables?
- A
Remove irrelevant columns such as 'Customer Notes' that do not contribute to predictions.
- B
Normalize numerical features to ensure they follow a standard distribution.
- C
Ensure categorical variables are encoded as strings or integers for AutoML Tables to process them.
- D
Manually fill missing values to prevent AutoML Tables from rejecting the dataset.
- E
Split the data into training, validation, and test sets before uploading it to AutoML Tables.
Show answer and explanation
Correct answers: A, C, E
Explanation
Preparing data for AutoML Tables involves removing irrelevant features to reduce noise, ensuring categorical variables are properly encoded, and splitting the dataset into training, validation, and test sets for robust evaluation. Steps like normalization and manual handling of missing values are unnecessary as AutoML Tables manages these automatically.
- A. Correct.
Correct: Removing irrelevant features like 'Customer Notes' is important to eliminate noise and improve model accuracy.
- B. Incorrect.
Incorrect: AutoML Tables automatically handles normalization and scaling of numerical features, so this step is unnecessary.
- C. Correct.
Correct: AutoML Tables requires categorical variables to be encoded appropriately as strings or integers for proper processing and feature engineering.
- D. Incorrect.
Incorrect: AutoML Tables can handle missing values and does not require them to be manually filled. It uses internal strategies to deal with missing data.
- E. Correct.
Correct: Splitting the data into training, validation, and test sets ensures proper evaluation and prevents data leakage during model training.