Google Professional Machine Learning Engineer Question 91
Select 3Google Cloud PlatformYou are a Machine Learning Engineer at a large retail company. Your team is tasked with building a recommendation system using historical sales data stored in BigQuery and product metadata stored in Cloud Storage. Before building the model, you need to preprocess the data and ensure it is clean, consistent, and ready for model training. Which of the following steps should you take to preprocess the data effectively?
- A
Use BigQuery SQL to standardize the sales data timestamps to a common format.
- B
Load product metadata from Cloud Storage into BigQuery and join it with the sales data for a unified dataset.
- C
Perform feature scaling and normalization directly in Cloud Storage before loading the data into BigQuery.
- D
Handle null values in the sales data using BigQuery's built-in SQL functions, such as COALESCE or IFNULL.
- E
Write a custom Python script to manually preprocess the data locally on your machine for better control over data quality.
Show answer and explanation
Correct answers: A, B, D
Explanation
Effective preprocessing of organization-wide data in Google Cloud involves leveraging the capabilities of BigQuery for data cleaning, transformation, and integration. This includes tasks like standardizing timestamps, handling null values, and combining datasets for a unified view. Preprocessing should utilize cloud-native tools to ensure scalability, reproducibility, and efficiency, avoiding manual or local approaches where possible.
- A. Correct.
Correct: Standardizing timestamps ensures consistency across the dataset, which is crucial for time-based analysis and modeling.
- B. Correct.
Correct: Joining product metadata with sales data provides a single, unified dataset that contains all necessary information for the recommendation system.
- C. Incorrect.
Incorrect: Feature scaling and normalization should be done after loading and combining the data, typically during the feature engineering phase, not directly in Cloud Storage.
- D. Correct.
Correct: Handling null values is an essential preprocessing step, and BigQuery SQL provides efficient built-in functions for this purpose.
- E. Incorrect.
Incorrect: Preprocessing data locally can lead to scalability issues and is not recommended for organization-wide data stored in cloud-based services like BigQuery and Cloud Storage.