acp module is outdated and does not work as it uses an older version of scikit-learn

[azedinez]

Please replace this bit:

nonconformist/nonconformist/acp.py

Lines 10 to 11 in cdef0d1

	from sklearn.cross_validation import KFold, StratifiedKFold
	from sklearn.cross_validation import ShuffleSplit, StratifiedShuffleSplit

by:
from sklearn.model_selection import KFold, StratifiedKFold
from sklearn.model_selection import ShuffleSplit, StratifiedShuffleSplit

[skaliev]

Do you know which version of scikit learn is stable with nonconformist 2.1.0 especially concerning acp?
I still get the following error (after editing acp.py):

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-2-944c165efcac> in <module>
     69 # -----------------------------------------------------------------------------
     70 for name, model in models.items():
---> 71     model.fit(data.data[train, :], data.target[train])
     72     prediction = model.predict(data.data[test, :], significance=significance)
     73     table = np.hstack((prediction, truth))

/zeus/fabiank/python3.7_env/lib/python3.7/site-packages/nonconformist/acp.py in fit(self, x, y)
    181                                                    self.n_models,
    182 		                                   problem_type)
--> 183                 for train, cal in samples:
    184                         predictor = clone(self.predictor)
    185                         predictor.fit(x[train, :], y[train])

/zeus/fabiank/python3.7_env/lib/python3.7/site-packages/nonconformist/acp.py in gen_samples(self, y, n_samples, problem_type)
     80 			splits = StratifiedShuffleSplit(y,
     81                                                         n_iter=n_samples,
---> 82 			                                test_size=self.cal_portion)
     83                 else:
     84 			splits = ShuffleSplit(y.size,

/zeus/fabiank/python3.7_env/lib/python3.7/site-packages/sklearn/utils/validation.py in inner_f(*args, **kwargs)
     70                           FutureWarning)
     71         kwargs.update({k: arg for k, arg in zip(sig.parameters, args)})
---> 72         return f(**kwargs)
     73     return inner_f
     74 

TypeError: __init__() got an unexpected keyword argument 'n_iter'

[juliameister]

Using scikit-learn 0.17 worked for me.

'n_iter' was renamed to 'n_splits' in the sklearn.model_selection.StratifiedShuffleSplit class in version 0.18 (see the relevant release history here: https://scikit-learn.org/0.18/whats_new.html).

[ben1a-hastings]

It seems this has been changed in the code, but this hasn't been updated on pypi. I am still getting the same error with the pypi version:

  File "C:\Users\auffarb\.conda\envs\py10\Lib\site-packages\nonconformist\acp.py", line 10, in <module>
    from sklearn.cross_validation import KFold, StratifiedKFold
ModuleNotFoundError: No module named 'sklearn.cross_validation'