def test_predict_classifier_wo_sigmoid(self):
ret_sql = f"""\
-- client: molehill/{molehill.__version__}
with features_exploded as (
select
id
, extract_feature(fv) as feature
, extract_weight(fv) as value
from
target_tbl t1
LATERAL VIEW explode(features) t2 as fv
)
-- DIGDAG_INSERT_LINE
select
t1.id
, sum(m1.weight * t1.value) as total_weight
from
features_exploded t1
left outer join model_tbl m1
on (t1.feature = m1.feature)
group by
t1.id
;
"""
pred_sql, pred_col = predict_classifier("target_tbl",
"id",
"model_tbl",
sigmoid=False)
assert pred_sql == ret_sql
assert pred_col == "total_weight"
def test_predict_classifier_bias_hashing(self):
ret_sql = f"""\
-- client: molehill/{molehill.__version__}
with features_exploded as (
select
id
, extract_feature(fv) as feature
, extract_weight(fv) as value
from
target_tbl t1
LATERAL VIEW explode(add_bias(feature_hashing(features))) t2 as fv
)
-- DIGDAG_INSERT_LINE
select
t1.id
, sigmoid(sum(m1.weight * t1.value)) as probability
from
features_exploded t1
left outer join model_tbl m1
on (t1.feature = m1.feature)
group by
t1.id
;
"""
pred_sql, pred_col = predict_classifier("target_tbl",
"id",
"model_tbl",
bias=True,
hashing=True)
assert pred_sql == ret_sql
assert pred_col == "probability"