In our regression tutorials, you can tackle realistic prediction problems by using several Hivemall's regression features such as:

• PA1a
• PA2a
• AROW
• AROWe

Our train_regressor function enables you to solve the regression problems with flexible configurable options. Let us try the function below.

It should be noted that the sample queries require you to prepare E2006-tfidf data. See our E2006-tfidf tutorial page for further instructions.

Training

create table e2006tfidf_regression_model as
select 
    feature,
    avg(weight) as weight
from (
    select 
      train_regressor(features,target,'-loss squaredloss -opt AdaGrad -reg no') as (feature,weight)
  from 
    e2006tfidf_train_x3
) t 
group by feature;

Prediction & evaluation

WITH predict as (
    select
      t.rowid, 
      sum(m.weight * t.value) as predicted
    from 
      e2006tfidf_test_exploded t LEFT OUTER JOIN
      e2006tfidf_regression_model m ON (t.feature = m.feature)
    group by
      t.rowid
),
submit as (
    select 
      t.target as actual, 
      p.predicted as predicted
    from 
      e2006tfidf_test t JOIN predict p 
        on (t.rowid = p.rowid)
)
select rmse(predicted, actual) from submit;