some questions about multi-class about mlbox HOT 5 CLOSED

Hi Axel,
Following the challenge meilleurdatascientistdefrance :-)
It seems multi-class detection is not obvious with the line
if (y_train.nunique() <= 2):
task = "classification"
in reader.py it limits to 2 class
For xgboost , objective and num_class params are also to be defined.
On local validation my score seems to be in first part of the LB
log_loss 0,96 , winners to 0,90
https://www.meilleurdatascientistdefrance.com/

Regards
Bruno

from mlbox.

Hello ! Thanks !
Unfortunately if your target is int64 it is considered as regression... see issue #54

from mlbox.

Thank you for.your reply,however,when I change the data’s type into int32 or int64,it doesn't works.what types should I change?

from mlbox.

Nice work @brunosez ! do you know many DS have tried MLBox during the challenge ?

from mlbox.

Hello, thanks for reporting this issue. In a next release, we will improve the automatic task detection and also add the possibility to override the task detection in case it is mis-detected...

from mlbox.