Evaluating the XGBoost classification model

To evaluate our BigQuery ML model, we'll use a ML.EVALUATE function and the table that we've expressly created as an evaluation dataset.

The following query will tell us if the model is suffering from overfitting or is able to also perform well on new data:

SELECT
  roc_auc,
  CASE
    WHEN roc_auc > .9 THEN 'EXCELLENT'
    WHEN roc_auc > .8 THEN 'VERY GOOD'
    WHEN roc_auc > .7 THEN 'GOOD'
    WHEN roc_auc > .6 THEN 'FINE'
    WHEN roc_auc > .5 THEN 'NEEDS IMPROVEMENTS'
  ELSE
  'POOR'
END
  AS model_quality
FROM 
  ML.EVALUATE(MODEL `10_nyc_trees_xgboost.xgboost_classification_model_version_3`,
    (
    SELECT
       ...