Google Professional Machine Learning Engineer Question 92
Select 3Google Cloud PlatformYour organization stores a large amount of data in Google Cloud Storage and BigQuery. You are tasked with building a machine learning model to predict customer churn. Before training the model, you need to preprocess the data to handle missing values, remove irrelevant features, and ensure that the data is ready for training. Which of the following steps are appropriate for exploring and preprocessing the data from these sources?
- A
Use BigQuery SQL queries to filter and aggregate relevant data before exporting it to a preprocessing pipeline.
- B
Directly train the machine learning model on raw data stored in Cloud Storage to save time.
- C
Leverage Dataflow to create a preprocessing pipeline for cleaning and transforming the data.
- D
Use Cloud Datalab or Vertex AI Workbench to visualize and analyze the data for patterns or anomalies.
- E
Ignore missing values in the data as most machine learning models can handle them directly.
Show answer and explanation
Correct answers: A, C, D
Explanation
Exploring and preprocessing organization-wide data involves steps like querying and filtering the data in BigQuery, using tools like Dataflow for scalable preprocessing, and leveraging visualization tools like Vertex AI Workbench to gain insights. These steps ensure the data is clean, relevant, and ready for model training. Ignoring preprocessing or missing values can negatively impact the model’s performance.
- A. Correct.
This is a correct step. BigQuery SQL queries allow you to efficiently filter, aggregate, and organize your data before exporting it for preprocessing, which is essential in the initial stages of data preparation.
- B. Incorrect.
This is incorrect. Training a machine learning model directly on raw data is not a best practice. Data should be preprocessed to clean and transform it into a suitable format for training.
- C. Correct.
This is a correct step. Dataflow is a powerful tool for building scalable preprocessing pipelines that can clean, transform, and prepare large datasets for machine learning.
- D. Correct.
This is a correct step. Cloud Datalab or Vertex AI Workbench provides tools for visualizing and exploring data, which helps identify patterns, anomalies, and preprocessing needs.
- E. Incorrect.
This is incorrect. Ignoring missing values can lead to poor model performance. Missing values should be handled explicitly, such as through imputation or removal, during preprocessing.