Catboost: Support pandas DataFrame type 'category'

[khrisanfov]

Pandas: 0.24.2
Catboost: 0.13.1

#!/usr/bin/env python3
import catboost
from catboost import datasets
from sklearn.model_selection import train_test_split

(train_df, test_df) = catboost.datasets.amazon()
y = train_df.ACTION
X = train_df.drop('ACTION', axis=1).astype('category')
# .astype('category') важно, если убрать то ошибки нет, кстати с .astype('category') Pool создается намного медленнее, почему?
cat_features = list(range(9))

# странно что так работает
pool = catboost.Pool(X, y, cat_features=cat_features)  # OK

# а после разбиения датасета ошибка
x_train, x_valid, y_train, y_valid = train_test_split(X, y, test_size=0.2)
train = catboost.Pool(x_train, y_train, cat_features=cat_features)  # ERROR
valid = catboost.Pool(x_valid, y_valid, cat_features=cat_features)  # ERROR

Traceback (most recent call last):
  File "/usr/local/Cellar/python3/3.6.3/Frameworks/Python.framework/Versions/3.6/lib/python3.6/site-packages/pandas/core/indexes/base.py", line 2657, in get_loc
    return self._engine.get_loc(key)
  File "pandas/_libs/index.pyx", line 108, in pandas._libs.index.IndexEngine.get_loc
  File "pandas/_libs/index.pyx", line 132, in pandas._libs.index.IndexEngine.get_loc
  File "pandas/_libs/hashtable_class_helper.pxi", line 987, in pandas._libs.hashtable.Int64HashTable.get_item
  File "pandas/_libs/hashtable_class_helper.pxi", line 993, in pandas._libs.hashtable.Int64HashTable.get_item
KeyError: 0

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/test.py", line 17, in <module>
    train = catboost.Pool(x_train, y_train, cat_features=cat_features)  # ERROR
  File "/usr/local/Cellar/python3/3.6.3/Frameworks/Python.framework/Versions/3.6/lib/python3.6/site-packages/catboost/core.py", line 291, in __init__
    self._init(data, label, cat_features, pairs, weight, group_id, group_weight, subgroup_id, pairs_weight, baseline, feature_names)
  File "/usr/local/Cellar/python3/3.6.3/Frameworks/Python.framework/Versions/3.6/lib/python3.6/site-packages/catboost/core.py", line 646, in _init
    self._init_pool(data, label, cat_features, pairs, weight, group_id, group_weight, subgroup_id, pairs_weight, baseline, feature_names)
  File "_catboost.pyx", line 1942, in _catboost._PoolBase._init_pool
  File "_catboost.pyx", line 1973, in _catboost._PoolBase._init_pool
  File "_catboost.pyx", line 1842, in _catboost._PoolBase._init_features_order_layout_pool
  File "_catboost.pyx", line 1489, in _catboost._set_features_order_data_pd_data_frame
  File "/usr/local/Cellar/python3/3.6.3/Frameworks/Python.framework/Versions/3.6/lib/python3.6/site-packages/pandas/core/series.py", line 868, in __getitem__
    result = self.index.get_value(self, key)
  File "/usr/local/Cellar/python3/3.6.3/Frameworks/Python.framework/Versions/3.6/lib/python3.6/site-packages/pandas/core/indexes/base.py", line 4360, in get_value
    iloc = self.get_loc(key)
  File "/usr/local/Cellar/python3/3.6.3/Frameworks/Python.framework/Versions/3.6/lib/python3.6/site-packages/pandas/core/indexes/base.py", line 2659, in get_loc
    return self._engine.get_loc(self._maybe_cast_indexer(key))
  File "pandas/_libs/index.pyx", line 108, in pandas._libs.index.IndexEngine.get_loc
  File "pandas/_libs/index.pyx", line 132, in pandas._libs.index.IndexEngine.get_loc
  File "pandas/_libs/hashtable_class_helper.pxi", line 987, in pandas._libs.hashtable.Int64HashTable.get_item
  File "pandas/_libs/hashtable_class_helper.pxi", line 993, in pandas._libs.hashtable.Int64HashTable.get_item
KeyError: 0

[indigoviolet]

Yes, I have a similar problem. It seems like there is some bad interaction with having pandas category columns in my dataframe:

• not passing cat_features will break with an error like _catboost.CatboostError: Bad value for num_feature[0,3]="GROCERY": Cannot convert ... to float

  • passing cat_features will produce a KeyError like above

However, if I leave those columns as string columns and pass cat_features in, CatBoost seems to work fine. BTW, I'm even able to pass cat_features in as string column names (which doesn't appear to be in the documentation yet)

[khrisanfov]

@indigoviolet I temporarily install previous version of pandas (0.23.4), it works fine for me.

[VeeraSrikanth24]

Just change the categorical variables to the string by .astype(str)
and when you check the data types by using .dtypes you will see them as "objects"
then pass the cat_features as a list of indices. That would work fine.

Don't change the data type to the category by .astype('category')

[andrey-khropov]

Has been fixed in CatBoost==0.16.