Pandas + Scikit learn: problem with stratified k-fold
When used with a StratifiedKFold
dataframe from scikit-learn, it returns a list of indices from 0 to n instead of a list of values from the DF index. Is there a way to change this?
Example:
df = pd.DataFrame()
df["test"] = (0, 1, 2, 3, 4, 5, 6)
df.index = ('a', 'b', 'c', 'd', 'e', 'f', 'g')
for i, (train, test) in enumerate(StratifiedKFold(df.index)):
print i, (train, test)
gives:
0 (array([], dtype=64), array([0,1,2,3,4,5,6])
1 (array([0,1,2,3,4,5,6]), array([], dtype=64))
2 (array([0,1,2,3,4,5,6]), array([], dtype=64))
I expect the index from df to be returned, not a range of length df ...
+3
source to share
1 answer
You only got the numbers df.index
that you selected StratifiedKFold
.
To change it back to the index of your DataFrame, simply
for i, (train, test) in enumerate(StratifiedKFold(df.index)):
print i, (df.index[train], df.index[test])
which gives
0 (Index([], dtype='object'), Index([u'a', u'b', u'c', u'd', u'e', u'f', u'g'], dtype='object'))
1 (Index([u'a', u'b', u'c', u'd', u'e', u'f', u'g'], dtype='object'), Index([], dtype='object'))
2 (Index([u'a', u'b', u'c', u'd', u'e', u'f', u'g'], dtype='object'), Index([], dtype='object'))
+3
source to share