How to fix 'name 'cross_validation' is not defined' error in python

Question

I am trying to run XGBClassifier parameter tuning and get a "'name 'cross_validation' is not defined" error following this line of code:

  kfold_5 =  cross_validation.KFold(n = len(X), shuffle = True, n_folds = numFolds)

Maybe I didn't import the appropriate library?

let me know if my answer helps – seralouk Jul 27 '19 at 15:52 — seralouk, Jul 27 '19 at 15:52
@serafeim great answer worked perfectly – Erez Ben-Moshe Jul 28 '19 at 11:56 — Erez Ben-Moshe, Jul 28 '19 at 11:56

seralouk · Accepted Answer · 2019-07-27T16:53:26.523

2

First, get your version:

import sklearn
sklearn.__version__

After scikit-learn version 0.17, the cross_validation.KFold has been migrated to model_selection.KFold.

If you have the 0.17 version, use this:

from sklearn.cross_validation import KFold

kfold_5 = KFold(n= len(X), n_folds = numFolds, shuffle=True)

If you have a version newer than 0.17, use this:

from sklearn.model_selection import KFold

kfold_5 = KFold(n_splits = numFolds, shuffle=True)

edited Jul 27 '19 at 16:53

answered Jul 27 '19 at 15:51

seralouk

The import statement depends on the OP's sklearn version. I have updated my answer. – seralouk Jul 27 '19 at 16:07
Indeed it is - (+1) now (previous versions of the answer assumed that OP had a new version, while such info is not provided) – desertnaut Jul 27 '19 at 16:07
@serafeim For the first code snippet I get "No module named 'sklearn.cross_validation' For the 2nd I get "TypeError: __init__() got an unexpected keyword argument 'n'" – Erez Ben-Moshe Jul 27 '19 at 16:33
for the second, did you type exactly `kfold_5 = KFold(n_splits = numFolds, shuffle=True)` ? The input arguments are not the same as in the first case, in this case. – seralouk Jul 27 '19 at 16:36
@ErezBen-Moshe please pay close attention to the provided snippets; in the 2nd snippet there should **not** be any argument `n` – desertnaut Jul 27 '19 at 17:57

1 Answers1