Google Professional Machine Learning Engineer Question 256
Select 3Google Cloud PlatformYour team is tasked with building a machine learning model to predict customer churn using a tabular dataset. The team wants to use Google Cloud’s AutoML Tables to train the model and deploy it on Kubernetes Engine for scalable predictions. Which of the following steps must be included in your workflow to achieve this?
- A
Upload the tabular dataset to a Google Cloud Storage bucket and create an AutoML Tables dataset.
- B
Use AutoML Tables to train the model and export the trained model as a Docker container.
- C
Deploy the exported model directly to Google App Engine for predictions.
- D
Use Kubernetes Engine to create a cluster and deploy the Docker image of the trained model for scalable predictions.
- E
Manually preprocess the tabular dataset before uploading it to AutoML Tables.
Show answer and explanation
Correct answers: A, B, D
Explanation
The correct workflow involves uploading the dataset to Google Cloud Storage for use with AutoML Tables, training the model, and exporting it as a Docker container. The containerized model can then be deployed on Kubernetes Engine for scalable predictions. App Engine is not relevant to this scenario, and manual preprocessing is unnecessary as AutoML Tables handles preprocessing automatically.
- A. Correct.
Correct: AutoML Tables requires the tabular dataset to be available in a Google Cloud Storage bucket to create a dataset for training.
- B. Correct.
Correct: AutoML Tables allows you to export a trained model as a Docker container, which can then be deployed on Kubernetes Engine.
- C. Incorrect.
Incorrect: While Google App Engine is a deployment option, the scenario specifies using Kubernetes Engine for scalable predictions.
- D. Correct.
Correct: Kubernetes Engine can be used to create a cluster and deploy the Docker container for scalable predictions.
- E. Incorrect.
Incorrect: AutoML Tables includes built-in preprocessing capabilities, so manual preprocessing is not required unless there are very specific preprocessing needs that AutoML cannot handle.