Combining Commands

SQL-inf supports combining SQL-inf commands through chaining.

For example, to use EXPLAIN you need to chain it with another command.

SELECT * FROM customers EXPLAIN(PREDICT(churn))

You can also chain several commands together through nesting of SQL-statements. In this example we nest a SENTIMENTstatement within an EXPLAIN-PREDICT statement to allow us to explain the inner command, ie the output of the SENTIMENT.

SELECT * FROM (SELECT * FROM customer_feedback SENTIMENT("Review Text")) EXPLAIN(PREDICT(prediction))

Finally, you can join multiple SQL-inf statements together to derive combined statistics. In this example (from the tutorial on Lead Scoring), two PREDICT statements are run and joined to calculate a weighted revenue prediction.

SELECT left_stmt.mql_id, 12 * conversion.prediction * revenue.prediction as expected_arr FROM (  
    (SELECT mql_id, landing_page_id, origin, declared_monthly_revenue, prediction 
        FROM Olist_Seller_Data PREDICT(declared_monthly_revenue, ignore=mql_id, model='reg')) as revenue
INNER JOIN 
    (SELECT mql_id, landing_page_id, origin, prediction,
        CASE WHEN seller_id IS NULL THEN 0 ELSE 1 END as converted
        FROM Olist_Seller_Data PREDICT(converted, ignore=mql_id)) as conversion
ON revenue.mql_id = conversion.mql_id
) WHERE converted=0 ORDER BY expected_arr DESC