sklearn.model_selection.train_test_split

数据集划分：sklearn.model_selection.train_test_split(*arrays, **options)

主要参数说明：

*arrays：可以是列表、numpy数组、scipy稀疏矩阵或pandas的数据框

test_size：可以为浮点、整数或None，默认为None

①若为浮点时，表示测试集占总样本的百分比

②若为整数时，表示测试样本样本数

③若为None时，test size自动设置成0.25

train_size：可以为浮点、整数或None，默认为None

①若为浮点时，表示训练集占总样本的百分比

②若为整数时，表示训练样本的样本数

③若为None时，train_size自动被设置成0.75

random_state：可以为整数、RandomState实例或None，默认为None

①若为None时，每次生成的数据都是随机，可能不一样

②若为整数时，每次生成的数据都相同

stratify：可以为类似数组或None

①若为None时，划分出来的测试集或训练集中，其类标签的比例也是随机的

②若不为None时，划分出来的测试集或训练集中，其类标签的比例同输入的数组中类标签的比例相同，可以用于处理不均衡的数据集

通过简单栗子看看各个参数的作用：

①test_size决定划分测试、训练集比例

In [1]: import numpy as np
   ...: from sklearn.model_selection import train_test_split
   ...: X = np.arange(20)
   ...: y = ['A','B','A','A','A','B','A','B','B','A','A','B','B','A','A','B','A
   ...: ','B','A','A']
   ...: X_train , X_test , y_train,y_test = train_test_split(X,y,test_size=0.25
   ...: ,random_state=0)
   ...:

In [2]: X_test.shape
Out[2]: (5,)

In [3]: X_train.shape
Out[3]: (15,)

In [4]: X_test ,y_test
Out[4]: (array([18,  1, 19,  8, 10]), ['A', 'B', 'A', 'B', 'A'])

②random_state不同值获取到不同的数据集

设置random_state=0再运行一次，结果同上述相同

In [5]: import numpy as np
   ...: from sklearn.model_selection import train_test_split
   ...: X = np.arange(20)
   ...: y = ['A','B','A','A','A','B','A','B','B','A','A','B','B','A','A','B','A
   ...: ','B','A','A']
   ...: X_train , X_test , y_train,y_test = train_test_split(X,y,test_size=0.25
   ...: ,random_state=0)
   ...: X_test ,y_test
   ...:
Out[5]: (array([18,  1, 19,  8, 10]), ['A', 'B', 'A', 'B', 'A'])

设置random_state=None运行两次，发现两次的结果不同

In [6]: import numpy as np
   ...: from sklearn.model_selection import train_test_split
   ...: X = np.arange(20)
   ...: y = ['A','B','A','A','A','B','A','B','B','A','A','B','B','A','A','B','A
   ...: ','B','A','A']
   ...: X_train , X_test , y_train,y_test = train_test_split(X,y,test_size=0.25
   ...: )
   ...: X_test ,y_test
   ...:
Out[6]: (array([ 3, 18, 14,  7,  4]), ['A', 'A', 'A', 'B', 'A'])

In [7]: import numpy as np
   ...: from sklearn.model_selection import train_test_split
   ...: X = np.arange(20)
   ...: y = ['A','B','A','A','A','B','A','B','B','A','A','B','B','A','A','B','A
   ...: ','B','A','A']
   ...: X_train , X_test , y_train,y_test = train_test_split(X,y,test_size=0.25
   ...: )
   ...: X_test ,y_test
   ...:
Out[7]: (array([18,  6,  3, 14,  8]), ['A', 'A', 'A', 'A', 'B'])

③设置stratify参数，可以处理数据不平衡问题

In [8]: import numpy as np
   ...: from sklearn.model_selection import train_test_split
   ...: X = np.arange(20)
   ...: y = ['A','B','A','A','A','B','A','B','B','A','A','B','B','A','A','B','A
   ...: ','B','A','A']
   ...: X_train , X_test , y_train,y_test = train_test_split(X,y,test_size=0.25
   ...: ,stratify=y)
   ...: X_test ,y_test
   ...:
Out[8]: (array([18,  8,  3, 10, 11]), ['A', 'B', 'A', 'A', 'B'])

In [9]: import numpy as np
   ...: from sklearn.model_selection import train_test_split
   ...: X = np.arange(20)
   ...: y = ['A','B','A','A','A','B','A','B','B','A','A','B','B','A','A','B','A
   ...: ','B','A','A']
   ...: X_train , X_test , y_train,y_test = train_test_split(X,y,test_size=0.25
   ...: ,stratify=y)
   ...: X_test ,y_test
   ...:
Out[9]: (array([ 6, 19,  8, 17,  0]), ['A', 'A', 'B', 'B', 'A'])

In [10]: X_train,y_train
Out[10]:
(array([ 7,  1, 11, 10, 15,  2,  3,  5,  4, 13, 12, 16, 18, 14,  9]),
 ['B', 'B', 'B', 'A', 'B', 'A', 'A', 'B', 'A', 'A', 'B', 'A', 'A', 'A', 'A'])

设置stratify=y时，我们发现每次划分后，测试集和训练集中的类标签比例同原始的样本中类标签的比例相同，都为2:3