Remove usage of scikit-learn's LabelEncoder
#3161
Conversation
Codecov Report
@@ Coverage Diff @@
## main #3161 +/- ##
=======================================
+ Coverage 99.7% 99.7% +0.1%
=======================================
Files 324 324
Lines 31233 31232 -1
=======================================
Hits 31128 31128
+ Misses 105 104 -1
Continue to review full report at Codecov.
|
angela97lin
requested review from
bchen1116,
christopherbunn and
jeremyliweishih
jeremyliweishih
left a comment
jeremyliweishih
left a comment
There was a problem hiding this comment.
Impeccable work!
| self._label_encoder = LabelEncoder() | ||
| y_encoded = pd.Series( | ||
| self._label_encoder.fit_transform(y_encoded), dtype="int64" | ||
| self._label_encoder.fit_transform(None, y_encoded)[1], dtype="int64" |
There was a problem hiding this comment.
do we need the pd.Series call here when we don't have it in catboost?
| if not is_integer_dtype(y): | ||
| self._label_encoder = LabelEncoder() | ||
| y = pd.Series(self._label_encoder.fit_transform(y), dtype="int64") | ||
| y = pd.Series(self._label_encoder.fit_transform(None, y)[1], dtype="int64") |
There was a problem hiding this comment.
Same question here!
bchen1116
left a comment
bchen1116
left a comment
There was a problem hiding this comment.
LGTM! Agreed with @jeremyliweishih, the label encoder component already returns the y value as an int-encoded series, so I don't think we need to do the extra initialization and type-casting in lightgbm and xgboost
Closes #3132