Citizen Data Scientists and Data Analysts who are comfortable with SQL query and don’t want to mess with other ML model building languages should prepare their ML Models using SQL queries.
Personally, I believe that there are a lot of folks who have their comfort level using SQL and get overwhelmed with anything new. Standard ANSI SQL can be used to build ML Models on BigQueryML.
Please check this video on my GCP channel for the steps to build ML Model using BigQueryML.
Here are the steps that I’ve used to build the BigQueryML model,
Step #1 – Load or utilize the existing data as your source dataset
Step #2 – Add a random value to create a split for test/train/evaluation within your dataset
Step #3 – Create an Input View with the labels of data records with Training, Evaluation, Prediction to be used for training and testing the source dataset
Step #4 – Build a model using the training dataset, in my case, I used Logistic Regression to solve the classification problem
Step #5 – Evaluate the ML Model
Step #6 – If the accuracy is good, and the model works fine, then use the model to predict the Prediction Dataset
Step #7 – Your Model is ready and is able to Predict. You can save this data and further explore in data studio
Please check BigQuery ML documentation from Google for further reading.