Google Professional Machine Learning Engineer Question 61
Select 3Google Cloud PlatformYou are tasked with preparing a dataset to train a model using AutoML Tables on Google Cloud. The dataset contains raw transactional data, including customer IDs, transaction timestamps, and purchase amounts. What steps should you take to ensure the dataset is ready for AutoML Tables?
- A
Remove customer IDs as they are unique identifiers and do not contribute to model training.
- B
Convert the transaction timestamps into meaningful features such as day of the week or time of the day.
- C
Manually scale the purchase amounts to a range between 0 and 1 to improve model accuracy.
- D
Upload the raw data to AutoML Tables directly, as it automatically performs all feature engineering.
- E
Ensure the data is labeled with a target column representing the outcome you want to predict.
Show answer and explanation
Correct answers: A, B, E
Explanation
To prepare data for AutoML Tables, it is important to remove irrelevant or non-predictive features such as unique IDs, derive meaningful features from raw data such as timestamps, and ensure the dataset includes a target column for prediction. While AutoML Tables automates many preprocessing tasks, some level of feature engineering and labeling is required from the user to ensure the data is in an appropriate format for model training.
- A. Correct.
Customer IDs are unique identifiers and do not provide predictive value for the model. They should be excluded from the dataset to improve model performance.
- B. Correct.
Timestamps in raw form are not directly useful for the model, but converting them into derived features such as 'day of the week' or 'time of day' can provide meaningful patterns for training.
- C. Incorrect.
Scaling is typically not required for AutoML Tables, as it handles feature preprocessing automatically. Manual scaling is unnecessary.
- D. Incorrect.
AutoML Tables does automate many preprocessing steps, but uploading raw data without preparing features or target labels will result in suboptimal performance or errors.
- E. Correct.
AutoML Tables requires a labeled dataset with a target column to predict. Without a target column, the training process cannot proceed.