training

-- SET mapred.reduce.tasks=32;
drop table kdd10b_arow_model1;
create table kdd10b_arow_model1 as
select 
 feature,
 -- voted_avg(weight) as weight
 argmin_kld(weight, covar) as weight -- [hivemall v0.2alpha3 or later]
from 
 (select 
     -- train_arow(addBias(features),label) as (feature,weight) -- [hivemall v0.1]
     train_arow(addBias(features),label) as (feature,weight,covar) -- [hivemall v0.2 or later]
  from 
     kdd10b_train_x3
 ) t 
group by feature;

prediction

create or replace view kdd10b_arow_predict1 
as
select
  t.rowid, 
  sum(m.weight * t.value) as total_weight,
  case when sum(m.weight * t.value) > 0.0 then 1 else -1 end as label
from 
  kdd10b_test_exploded t LEFT OUTER JOIN
  kdd10b_arow_model1 m ON (t.feature = m.feature)
group by
  t.rowid;

evaluation

create or replace view kdd10b_arow_submit1 as
select 
  t.rowid, 
  t.label as actual, 
  pd.label as predicted
from 
  kdd10b_test t JOIN kdd10b_arow_predict1 pd 
    on (t.rowid = pd.rowid);
select count(1)/748401 from kdd10b_arow_submit1 
where actual = predicted;

0.8565808971393678

results matching ""

    No results matching ""