for Imbalanced data dealing with cat boost

Question

Is there a parameter like "scale_pos_weight" in catboost package as we used to have in the xgboost package in python ?

score 11 · Accepted Answer · answered Aug 16 '17 at 08:41

11

Yes, the parameter is named "class_weights", you can find it here : Documentation

You have to pass a list like [0.8, 0.2] for binary or [0.3, 0.8, 0.4, 0.6] for multiclass of 4 for example. Doesn't have to sum to 1, it's used as a multiplier.

answered Aug 16 '17 at 08:41

Gaarv

814
8
15

score 8 · Answer 2 · answered Jul 05 '18 at 14:22

8

CatBoost also has scale_pos_weight parameter starting from version 0.6.1

answered Jul 05 '18 at 14:22

Anna Veronika Dorogush

1,103
10
10

Can you provide some more information (ex a link or an example?) – kosnik Jul 05 '18 at 14:45
@Kosnik You can check here https://tech.yandex.com/catboost/doc/dg/concepts/python-reference_catboostclassifier-docpage/ – Harshit Mehta Jul 11 '18 at 02:22

score 2 · Answer 3 · answered Mar 06 '21 at 11:06

2

Yes, you can use "auto_class_weights" and "scale_pos_weight" as described in original catboost documentations: https://catboost.ai/docs/concepts/python-reference_parameters-list.html

answered Mar 06 '21 at 11:06

from_mars

31
3

for Imbalanced data dealing with cat boost

3 Answers3