BigQuery Machine Learning (BQML) is a BigQuery’s feature that allows users to create and execute machine learning using SQL. As BQML is distinctive from standard SQL queries, implementation of BQML in Looker requires additional processes and LookML parameters that are completely different from a traditional Looker data modeling workflow.
I developed a tutorial covering step-by-step instruction on how to implement BQML with Looker syntax. Through this tutorial, we can:
- Understand the different technicalities and syntaxes between BigQuery and Looker (i.e.: “model table” for BQML vs. Looker’s persistent derived tables, LookML parameters for triggering caching policies, triggering table build, applying BigQuery’s “model table” with a standard Looker view to make predictions, etc)
- Implement a data workflow following a standard 5-step machine learning method (i.e.: prepare data, cleaning data, making models, evaluating models, making predictions, etc)
The tutorial is using the BigQuery free public dataset so anyone with a GCP account and a Looker instance can start testing and implementing this feature quickly.
Hope this helps
- LookML BigQuery Block
- BigQuery Machine Learning Coursera