KeyError：“[Int64Index dtype='int64', length=9313)] 都不在 [columns]”

Question

have a dataframe of 323 column and 10348 row . i want to divide it using stratified k-Fold using the following code

df= pd.read_csv("path")
 x=df.loc[:, ~df.columns.isin(['flag'])]
 y= df['flag']
StratifiedKFold(n_splits=5, random_state=None, shuffle=False)
for train_index, test_index in skf.split(x, y):
       print("TRAIN:", train_index, "TEST:", test_index)
       x_train, x_test = x[train_index], x[test_index]
       y_train, y_test = y[train_index], y[test_index]

but i get the following error

KeyError: "None of [Int64Index([    0,     1,     2,     3,     4,     5,     6,     7,     8,\n               10,\n            ...\n            10338, 10339, 10340, 10341, 10342, 10343, 10344, 10345, 10346,\n            10347],\n           dtype='int64', length=9313)] are in the [columns]"

any one tell me why i get this error and how to fix it

Answer 1

Seems like you have a data frame slicing issue rather than something wrong with StratifiedKFold itself. I crafted a df for that purpose and solved it using iloc to slice an array of indexes here:

from sklearn import model_selection

# The list of some column names in flag
flag = ["raw_sentence", "score"]
x=df.loc[:, ~df.columns.isin(flag)].copy()
y= df[flag].copy()
skf =model_selection.StratifiedKFold(n_splits=2, random_state=None, shuffle=False)
for train_index, test_index in skf.split(x, y):
    print("TRAIN:", train_index, "TEST:", test_index)
    x_train, x_test = x.iloc[list(train_index)], x.iloc[list(test_index)]

And train_indexes and test_indexes being nd-arrays kinda messes the work here, i convert them to the lists.

you may refer: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html

Answer 2

you can also use df.take(indices_list,axis=0)

x_train, x_test = x.take(list(train_index),axis=0), x.take(list(test_index),axis=0)

https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.take.html

Answer 3

Try changing pandas dataframe to numpy array as follow:

pd.DataFrame({"A": [1, 2], "B": [3, 4]}).to_numpy()

array([[1, 3],
       [2, 4]])