Add dictionary support for Undersampler base class #2235
Conversation
Codecov Report
@@ Coverage Diff @@
## main #2235 +/- ##
=======================================
+ Coverage 99.9% 99.9% +0.1%
=======================================
Files 280 280
Lines 24294 24330 +36
=======================================
+ Hits 24266 24302 +36
Misses 28 28
Continue to review full report at Codecov.
|
bchen1116
requested review from
ParthivNaresh,
angela97lin,
chukarsten,
dsherry,
freddyaboulton and
jeremyliweishih
| y_dict = y.value_counts().to_dict() | ||
| new_dic = {} | ||
| for k, v in self.sampling_ratio_dict.items(): | ||
| new_dic[k] = max(y_dict[k] - v, 0) |
There was a problem hiding this comment.
@bchen1116 The doc said that we'd accept ratios as the values of the dictionary but it seems like we're still expecting counts. Is that coming in the part 2 pr?
There was a problem hiding this comment.
Yep! I plan on adding support for converting the ratios to the number of samples in the base_sampler class so both the over and undersamplers have access to it
There was a problem hiding this comment.
Got it. I think I understand the plan now!
freddyaboulton
left a comment
freddyaboulton
left a comment
There was a problem hiding this comment.
I think this looks great @bchen1116 !
angela97lin
left a comment
angela97lin
left a comment
There was a problem hiding this comment.
LGTM! Left nit-picky comment about docstring and how sampling_ratio interacts with sampling_ratio_dict but otherwise 🚢
evalml/preprocessing/data_splitters/balanced_classification_sampler.py
Outdated
Show resolved
Hide resolved
3 participants
Part 1 of #2105
This PR adds support for accepting dictionary input in the undersampler base class. I will add support for adding the dictionary support for Undersampler in a separate PR, since I want to handle the dictionary processing/conversion from ratios to number of samples in the sampler base class, which requires this doc to be approved and implemented.