includes the STREET_ADDRESS infoType.
Explanation:
To retrieve all occurrences of street addresses from a BigQuery dataset, the most effective and comprehensive method is to use Cloud Data Loss Prevention (DLP). Here's why option A is the best choice:
Cloud Data Loss Prevention (DLP):
Cloud DLP is designed to discover, classify, and protect sensitive information. It includes pre-defined infoTypes for various kinds of sensitive data, including street addresses.
Using Cloud DLP ensures thorough and accurate detection of street addresses based on advanced pattern recognition and contextual analysis.
Deep Inspection Job:
A deep inspection job allows you to scan entire tables for sensitive information.
By creating an inspection template that includes the STREET_ADDRESS infoType, you can ensure that all instances of street addresses are detected across your dataset.
Scalability and Accuracy:
Cloud DLP is scalable and can handle large datasets efficiently.
It provides a high level of accuracy in identifying sensitive data, reducing the risk of missing any occurrences.
Steps to Implement:
Set Up Cloud DLP:
Enable the Cloud DLP API in your Google Cloud project.
Create an Inspection Template:
Create an inspection template in Cloud DLP that includes the STREET_ADDRESS infoType.
Run Deep Inspection Jobs:
Create and run a deep inspection job for each table in your dataset using the inspection template.
Review the inspection job results to retrieve all occurrences of street addresses.
Reference:
Cloud DLP Documentation
Creating Inspection Jobs

To resolve the concurrency issue in BigQuery caused by the introduction of hundreds of non-time-sensitive SQL pipelines, the best approach is to differentiate the types of queries based on their urgency and resource requirements. Here's why option C is the best choice:
SQL Pipelines as Batch Queries:
Batch queries in BigQuery are designed for non-time-sensitive operations. They run in a lower priority queue and do not consume slots immediately, which helps to reduce the overall slot consumption during peak times.
By converting non-time-sensitive SQL pipelines to batch queries, you can significantly alleviate the pressure on slot reservations.
Ad-Hoc Queries as Interactive Queries:
Interactive queries are prioritized to run immediately and are suitable for ad-hoc analysis where users expect quick results.
Running ad-hoc queries as interactive jobs ensures that analysts can get their results without delay, improving productivity and user satisfaction.
Concurrency Management:
This approach helps balance the workload by leveraging BigQuery's ability to handle different types of queries efficiently, reducing the likelihood of encountering quota errors due to slot exhaustion.
Steps to Implement:
Identify Non-Time-Sensitive Pipelines:
Review and identify SQL pipelines that are not time-critical and can be executed as batch jobs.
Update Pipelines to Batch Queries:
Modify these pipelines to run as batch queries. This can be done by setting the priority of the query job to BATCH.
Ensure Ad-Hoc Queries are Interactive:
Ensure that all ad-hoc queries are submitted as interactive jobs, allowing them to run with higher priority and immediate slot allocation.
Reference:
BigQuery Batch Queries
BigQuery Slot Allocation and Management

If you know the set of all possible feature values of a column and there are only a few of them, you can use categorical_column_with_vocabulary_list. Each key in the list will get assigned an auto-incremental ID starting from 0.
What if we don't know the set of possible values in advance? Not a problem. We can use categorical_column_with_hash_bucket instead. What will happen is that each possible value in the feature column occupation will be hashed to an integer ID as we encounter them in training.