Google Professional Machine Learning Engineer Question 93
Select 3Google Cloud PlatformYour organization stores customer data in Google Cloud Storage and transactional data in BigQuery. You need to preprocess and explore this data to prepare it for a machine learning model. The customer data contains unstructured text fields, while the transactional data is structured and includes null values. Which of the following steps should you take to ensure the data is ready for training?
- A
Use BigQuery to identify and replace null values in the transactional data with appropriate default values or statistical measures.
- B
Transform unstructured text fields in the customer data into numerical representations using techniques like tokenization or embedding.
- C
Directly join the customer data from Cloud Storage and transactional data from BigQuery without any preprocessing.
- D
Apply data validation checks to ensure the schemas of the two datasets are compatible before merging.
- E
Use Cloud Storage to automatically clean and preprocess both datasets without additional tools or scripts.
Show answer and explanation
Correct answers: A, B, D
Explanation
To prepare organization-wide data for machine learning, it is crucial to preprocess the data to handle issues like null values, schema mismatches, and unstructured fields. BigQuery provides powerful tools for managing structured data, while unstructured text fields from Cloud Storage require transformation using techniques like embeddings. Additionally, ensuring schema compatibility before merging datasets is a critical preprocessing step.
- A. Correct.
Correct: Handling null values is an essential preprocessing step to ensure the data is consistent and suitable for machine learning training. BigQuery is well-suited for these operations.
- B. Correct.
Correct: Text fields need to be transformed into numerical formats to be used in machine learning models. Techniques like tokenization or embeddings are commonly used for this purpose.
- C. Incorrect.
Incorrect: Directly joining datasets without preprocessing can lead to issues such as schema mismatches, null values, or unprocessed text fields, which can negatively impact the model.
- D. Correct.
Correct: Ensuring the schemas of the datasets are compatible is important before merging to avoid data alignment issues.
- E. Incorrect.
Incorrect: Cloud Storage is a storage solution and does not inherently provide preprocessing or data cleaning capabilities. Additional tools or scripts are required to preprocess data stored in Cloud Storage.