Doing inference with BQML
In supervised ML, the ultimate goal is to use a trained model to make predictions on new data. BQML provides the ML.PREDICT function for this purpose. Using this function, you can easily predict outcomes by supplying new data to a trained model. The ML.PREDICT function can be used during model creation, after model creation, or after a failure, so long as at least one iteration has been completed. The function returns a table with the same number of rows as the input table, and it includes all columns from the input table and all output columns from the model, with the output column names prefixed with predicted_.
ML.PREDICT(MODEL model_name,
{TABLE table_name | (query_statement)}
[, STRUCT<threshold FLOAT64,
keep_original_columns BOOL> settings)]) 
