Categorical variables cause NaN values.

[JelkeW]

Hi,
I had an issue with the resampled data containing lots of NaN values and thus SMOGN not running.
For anyone who is familiar with it: oops! synthetic data contains missing values
During debugging I figured out, that the NaN values only occur on categorical variables.

Two fixes for anyone encountering the problem:

1. Fix on data side
   Change all the column in your dataframe from type category to type object
data[column] = data[column].astype("object")
2. Fix on SMOGN side
   In smogn.over_sampling change
   nom_dtypes = ["object", "bool", "datetime64"]
   to
   nom_dtypes = ["object", "bool", "datetime64", "category"]

Took me a bit of time to figure it out. Hope it helps 😊