allow outlier trimmer and winsorizer to select optimal fold automatically based on the method to detect outliers

feature-engine / feature_engine

Feature engineering package with sklearn like functionality

https://feature-engine.trainindata.com/

BSD 3-Clause "New" or "Revised" License

1.79k stars 304 forks source link

Closed solegalli closed 1 month ago

solegalli commented 2 months ago

At the moment, the parameter fold from OutlierTrimmer and Winsorizer are set to 3.

I'd like it to be set to "auto" which would have this behaviour:

if iqr: then fold = 1.5
if gaussian, then fold=3
if MAD, then fold=3.29 (coincides with r package: https://easystats.github.io/performance/articles/check_outliers
if quantile: then 0.05

Easy fix, plus tests.

solegalli commented 2 months ago

@glevv could I tempt you with this one?

glevv commented 2 months ago

@glevv could I tempt you with this one?

Will look into it