CODEWITHSUNDEEP
scikit-learncross-validation
user11251362
user11251362

Reputation: 101

What changes in parameters should make when import StratifiedShuffleSplit from sklearn.model_selection by replacing sklearn.cross_validation

I was trying to run a python3 code for isolated speech recognition, there I got a DeprecationWarning for using:

from sklearn.cross_validation import StratifiedShuffleSplit

In order to remove this warning, I just imported StratifiedShuffleSplit from sklearn.model_selection instead of sklearn.cross_validation and after running the code, I got:

TypeError: 'StratifiedShuffleSplit' object is not iterable

Maybe because in

class sklearn.cross_validation.StratifiedShuffleSplit(y, n_iter=10, test_size=0.1, train_size=None, random_state=None)

y is an array.

While in:

class sklearn.cross_validation.StratifiedShuffleSplit(y, n_iter=10, test_size=0.1, train_size=None, random_state=None)

There is no array:

from sklearn.model_selection import StratifiedShuffleSplit
sss = StratifiedShuffleSplit(all_labels, test_size=0.1, random_state=0)

for n,i in enumerate(all_obs):
    all_obs[n] /= all_obs[n].sum(axis=0)

for train_index, test_index in sss:
    X_train, X_test = all_obs[train_index, ...], all_obs[test_index, ...]
    y_train, y_test = all_labels[train_index], all_labels[test_index]
ys = set(all_labels)
ms = [gmmhmm(7) for y in ys]

How to replace all_labels because it is an array according to sklearn.cross_validation but sklearn.model_selection don't accept an array parameter.

Upvotes: 1

Views: 551

Answers (1)

Alessandro Solbiati
Alessandro Solbiati

Reputation: 979

The difference between the two is that

  • sklearn.model_selection.StratifiedShuffleSplit is a cross-validator

  • sklearn.cross_validation.StratifiedShuffleSplit is a cross-validator iterator

Hence the correct usages from your example would be

from sklearn.model_selection import StratifiedShuffleSplit
sss = StratifiedShuffleSplit(test_size=0.1, random_state=0)

for n,i in enumerate(all_obs):
    all_obs[n] /= all_obs[n].sum(axis=0)

for train_index, test_index in sss.split(all_obs, all_labels):
     print(train_index, test_index)

It might be helpful to read the example in the docs of sklearn.model_selection.StratifiedShuffleSplit

Upvotes: 0

Related Questions