sklearn.model_selection.GroupKFold

グループKの交差検証:sklearn.model_selection.GroupKFold(n_splits=3)
パラメータの説明:
n_splits:折り返し、デフォルトは3、少なくとも2
注意点:同じグループのサンプルが同じ折り返しのテストセットとトレーニングセットに同時に現れることはありません.
①データセットは均等であり、各グループのサンプルも均等である

In [11]: from sklearn.model_selection import GroupKFold
    ...: import numpy as np
    ...: X = np.arange(24).reshape(12,2)
    ...: y = np.array([1,1,2,3,1,2,3,2,2,3,3,1])
    ...: groups = np.array([1,2,3,4,5,6,1,2,3,4,5,6])
    ...: kf = GroupKFold(n_splits=6)
    ...: for train_index , test_index in kf.split(X,y,groups):
    ...:     print('train_index:%s , test_index: %s ' %(train_index,test_index)
    ...: )
    ...:     print('train_groups:%s , test_groups: %s ' %(groups[train_index],g
    ...: roups[test_index]))
    ...:
train_index:[ 0  1  2  3  4  6  7  8  9 10] , test_index: [ 5 11]
train_groups:[1 2 3 4 5 1 2 3 4 5] , test_groups: [6 6]
train_index:[ 0  1  2  3  5  6  7  8  9 11] , test_index: [ 4 10]
train_groups:[1 2 3 4 6 1 2 3 4 6] , test_groups: [5 5]
train_index:[ 0  1  2  4  5  6  7  8 10 11] , test_index: [3 9]
train_groups:[1 2 3 5 6 1 2 3 5 6] , test_groups: [4 4]
train_index:[ 0  1  3  4  5  6  7  9 10 11] , test_index: [2 8]
train_groups:[1 2 4 5 6 1 2 4 5 6] , test_groups: [3 3]
train_index:[ 0  2  3  4  5  6  8  9 10 11] , test_index: [1 7]
train_groups:[1 3 4 5 6 1 3 4 5 6] , test_groups: [2 2]
train_index:[ 1  2  3  4  5  7  8  9 10 11] , test_index: [0 6]
train_groups:[2 3 4 5 6 2 3 4 5 6] , test_groups: [1 1]

②折数均等、組数不平衡

In [13]: #sklearn.model_selection.GroupKFold(n_splits=3)
    ...: from sklearn.model_selection import GroupKFold
    ...: import numpy as np
    ...: X = np.arange(24).reshape(12,2)
    ...: y = np.array([1,1,2,3,1,2,3,2,2,3,3,1])
    ...: groups = np.array([1,2,3,4,5,6,1,2,3,4,5,7])
    ...: kf = GroupKFold(n_splits=4)
    ...: for train_index , test_index in kf.split(X,y,groups):
    ...:     print('train_index:%s , test_index: %s ' %(train_index,test_index)
    ...: )
    ...:     print('train_groups:%s , test_groups: %s ' %(groups[train_index],g
    ...: roups[test_index]))
    ...:
train_index:[ 1  2  3  5  7  8  9 11] , test_index: [ 0  4  6 10]
train_groups:[2 3 4 6 2 3 4 7] , test_groups: [1 5 1 5]
train_index:[ 0  1  2  4  5  6  7  8 10] , test_index: [ 3  9 11]
train_groups:[1 2 3 5 6 1 2 3 5] , test_groups: [4 4 7]
train_index:[ 0  1  3  4  6  7  9 10 11] , test_index: [2 5 8]
train_groups:[1 2 4 5 1 2 4 5 7] , test_groups: [3 6 3]
train_index:[ 0  2  3  4  5  6  8  9 10 11] , test_index: [1 7]
train_groups:[1 3 4 5 6 1 3 4 5 7] , test_groups: [2 2]

③割引数が不均等で、グループ数が不均衡である

In [14]: from sklearn.model_selection import GroupKFold
    ...: import numpy as np
    ...: X = np.arange(24).reshape(12,2)
    ...: y = np.array([1,1,2,3,1,2,3,2,2,3,3,1])
    ...: groups = np.array([1,2,3,4,5,6,1,2,3,4,5,3])
    ...: kf = GroupKFold(n_splits=5)
    ...: for train_index , test_index in kf.split(X,y,groups):
    ...:     print('train_index:%s , test_index: %s ' %(train_index,test_index)
    ...: )
    ...:     print('train_groups:%s , test_groups: %s ' %(groups[train_index],g
    ...: roups[test_index]))
    ...:
train_index:[ 0  1  3  4  5  6  7  9 10] , test_index: [ 2  8 11]
train_groups:[1 2 4 5 6 1 2 4 5] , test_groups: [3 3 3]
train_index:[ 0  1  2  3  6  7  8  9 11] , test_index: [ 4  5 10]
train_groups:[1 2 3 4 1 2 3 4 3] , test_groups: [5 6 5]
train_index:[ 0  1  2  4  5  6  7  8 10 11] , test_index: [3 9]
train_groups:[1 2 3 5 6 1 2 3 5 3] , test_groups: [4 4]
train_index:[ 0  2  3  4  5  6  8  9 10 11] , test_index: [1 7]
train_groups:[1 3 4 5 6 1 3 4 5 3] , test_groups: [2 2]
train_index:[ 1  2  3  4  5  7  8  9 10 11] , test_index: [0 6]
train_groups:[2 3 4 5 6 2 3 4 5 3] , test_groups: [1 1]