FIX: SMOTENC should use half of the median of the std. dev.

[glemaitre]

bug fix in SMOTE-NC

[pep8speaks]

Hello @glemaitre! Thanks for submitting the PR.

• There are no PEP8 issues in the file imblearn/over_sampling/_smote.py !

[Eugene1518]

Hi. I am confused with the way of SmoteNC solving categorical variables. Could you tell me the difference between using median of the std. dev, and using one hot encoder to transform categorical feature into continuous feature with Smote. If I use one hot encoder with Smote instead of SmoteNC, what's the drawback？ Thanks very much

[glemaitre]

It is the way this is proposed in the original paper to which you should refer for more detailed explanations. Be aware that OneHotEncoder does not transform a categorical feature into a continuous feature but rather encodes a categorical feature into features that do not impose any order.

[Eugene1518]

thanks~