Dictionary support for Undersampler #2271
Conversation
Codecov Report
@@ Coverage Diff @@
## main #2271 +/- ##
=========================================
+ Coverage 100.0% 100.0% +0.1%
=========================================
Files 280 280
Lines 24336 24392 +56
=========================================
+ Hits 24314 24370 +56
Misses 22 22
Continue to review full report at Codecov.
|
bchen1116
requested review from
chukarsten,
dsherry,
freddyaboulton and
jeremyliweishih
There was a problem hiding this comment.
This is good work, Bryan. I think I need some chat with what this means to have a ratio for both the majority and minority class. That's a little confusing to me now that I'm looking at it. I think we also need to drop an update to the docstring and figure out what _dictionary_to_params() should be doing without 'sampling_ratio_dict' defined in the params.
evalml/pipelines/components/transformers/samplers/base_sampler.py
Outdated
Show resolved
Hide resolved
| @@ -59,6 +59,52 @@ def transform(self, X, y=None): | |||
| y = infer_feature_types(y) | |||
| return X, None | |||
|
|
|||
| def _convert_dictionary(self, sampling_dict, y): | |||
There was a problem hiding this comment.
Should this method live in the base class? It would be different for undersamplers vs oversampler right?
There was a problem hiding this comment.
I plan on doing an if-statement in the base class to handle the differences between over and undersamplers! Should be a one-line difference
fix #2105