Using BigQuery ML for online predictions

In this section, we'll understand how we can use a BigQuery ML model in a synchronous and online manner.

BigQuery ML represents a huge opportunity to democratize ML techniques for business and data analysts. When BigQuery ML is trained and ready to use, we can invoke it directly in BigQuery using a SQL query or we can export it into TensorFlow format.

The requirements of each use case drive the prediction type that we should adopt, as outlined here:

• We use online prediction when we want to enable request-response applications and when getting an immediate prediction is critical.
• We adopt batch prediction to process large volumes of data when we don't need immediate predictions—for example, scheduling daily or weekly jobs that calculate predictions on the data collected since the last job execution.

While using BigQuery SQL statements is more suitable for batch predictions on a large number of records...