Evaluating the multiclass logistic regression model

In this section, we'll execute queries to check the performance of the multiclass logistic regression model.

For the evaluation phase of our BigQuery ML model, we'll use the ML.EVALUATE function and the evaluation_table table, expressly created to host the evaluation records.

As we can see, the evaluation is performed on the same fields that were used during the training phase of the model but are extracted from the evaluation_table table that was created completely disjoint from the training dataset.

The external SELECT statement extracts the roc_auc value returned by the ML.EVALUATE function. It also provides a meaningful description of the quality of the model that starts from 'POOR' and goes up to the 'EXCELLENT' grade, passing through some intermediate stages such as 'NEEDS IMPROVEMENTS' and 'GOOD'.

Let's execute the following query to extract the key performance...