bug de sklearn KhiopsClassifier quand on utilise un numpy.ndarray en X

[n-voisine]

Il semble y avoir un bug quand on utilise un numpy.ndarray pour X avec des variables continue et catégoriel
Ça passe si on utilise que des continue

code testé sur Adult
print("Type de la variable :", type(X_current))
print("Shape (dimensions) :", X_current.shape)
print("Type de données :", X_current.dtype)
print("Premier exemple (première ligne) :", X_current[0])
print("Aperçu du tableau :\n", X_current)
from khiops.sklearn import KhiopsClassifier
khc = KhiopsClassifier()
khc.fit(X_current, y)

retour
Type de la variable : <class 'numpy.ndarray'>
Shape (dimensions) : (39073, 15)
Type de données : object
Premier exemple (première ligne) : [34496 37 'Private' 193106 'Bachelors' 13 'Never-married' 'Sales'
'Not-in-family' 'White' 'Female' 0 0 30 'United-States']
Aperçu du tableau :
[[34496 37 'Private' ... 0 30 'United-States']
[18592 56 'Self-emp-inc' ... 1651 40 'United-States']
[12563 53 'Private' ... 0 35 'United-States']
...
[26499 61 'Federal-gov' ... 1887 40 'United-States']
[20648 38 'Private' ... 0 40 'United-States']
[42716 43 'State-gov' ... 0 40 'United-States']]
✗ Error: could not convert string to float: 'Private'
Traceback (most recent call last):
File "C:\DEV\boosting\bench_classification.py", line 346, in benchmark_classification
results = evaluate_model(
^^^^^^^^^^^^^^^
File "C:\DEV\boosting\bench_classification.py", line 207, in evaluate_model
model.fit(X_train, y_train)
File "C:\DEV\boosting\rfkhiops.py", line 146, in fit
khc.fit(X_current, y)
File "C:\Users\khiops11_bt\Lib\site-packages\khiops\sklearn\estimators.py", line 1863, in fit
return super().fit(X, y, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\khiops11_bt\Lib\site-packages\khiops\sklearn\estimators.py", line 1262, in fit
super().fit(X, y=y, **kwargs)
File "C:\Users\khiops11_bt\Lib\site-packages\khiops\sklearn\estimators.py", line 369, in fit
dataset = Dataset(X, y, categorical_target=categorical_target)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\khiops11_bt\Lib\site-packages\khiops\sklearn\dataset.py", line 441, in init
self.main_table = NumpyTable("main_table", X)
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\khiops11_bt\Lib\site-packages\khiops\sklearn\dataset.py", line 1080, in init
self.data_source = check_array(array, ensure_2d=True, force_all_finite=False)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\khiops11_bt\Lib\site-packages\sklearn\utils\validation.py", line 1053, in check_array
array = _asarray_with_order(array, order=order, dtype=dtype, xp=xp)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\khiops11_bt\Lib\site-packages\sklearn\utils_array_api.py", line 757, in _asarray_with_order
array = numpy.asarray(array, order=order, dtype=dtype)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: could not convert string to float: 'Private'

[folmos-at-orange]

When having a Numpy array as input, a scikit-learn estimator expects it to be numerical (your example should fail any other scikit-learn classifier). If you want to use string columns simply use a pandas dataframe.