如何把数据集划分成训练集和测试集

一、留出法(hold-out)

通过调用sklearn.model_selection.train_test_split按比例划分训练集和测试集:

import numpy as np
from sklearn.model_selection import train_test_split

X, Y = np.arange(10).reshape((5, 2)), range(5)
print("X=", X)
print("Y=", Y)
X_train, X_test, Y_train, Y_test = train_test_split(
    X, Y, test_size=0.30, random_state=42)
print("X_train=", X_train)
print("X_test=", X_test)
print("Y_train=", Y_train)
print("Y_test=", Y_test)

其中test_size=0.30表示T占30%， 那么S占70%。运行结果：

X= [[0 1]
 [2 3]
 [4 5]
 [6 7]
 [8 9]]
Y= range(0, 5)
X_train= [[4 5]
 [0 1]
 [6 7]]
X_test= [[2 3]
 [8 9]]
Y_train= [2, 0, 3]
Y_test= [1, 4]

二、交叉验证法(cross validation)

通过调用sklearn.model_selection.KFold按k折交叉划分训练集和测试集:

import numpy as np
from sklearn.model_selection import KFold

X= np.arange(10).reshape((5, 2))
print("X=", X)
kf = KFold(n_splits=2)
for train_index, test_index in kf.split(X):
    print('X_train:%s ' % X[train_index])
    print('X_test: %s ' % X[test_index])

其中n_splits=2表示k=2。运行结果：

X= [[0 1]
 [2 3]
 [4 5]
 [6 7]
 [8 9]]
X_train:[[6 7]
 [8 9]] 
X_test: [[0 1]
 [2 3]
 [4 5]] 
X_train:[[0 1]
 [2 3]
 [4 5]] 
X_test: [[6 7]
 [8 9]]  

当D包含m个样本，令k=m，则得到交叉验证法的一个特例——留一法(Leave-One-Out，简称LOO)。留一法使用的S只比D少一个样本，所以在绝大多数情况下，实际评估结果与用D训练的模型相似。因此，留一法被认为比较准确。但留一法对于大数据集，计算开销太大；另外也不见得永远比其他方法准确。
通过调用sklearn.model_selection.LeaveOneOut按留一法划分训练集和测试集:

import numpy as np
from sklearn.model_selection import LeaveOneOut

X= np.arange(10).reshape((5, 2))
print("X=", X)
loo = LeaveOneOut()
for train_index, test_index in loo.split(X):
    print('X_train:%s ' % X[train_index])
    print('X_test: %s ' % X[test_index])

运行结果：

X= [[0 1]
 [2 3]
 [4 5]
 [6 7]
 [8 9]]
X_train:[[2 3]
 [4 5]
 [6 7]
 [8 9]] 
X_test: [[0 1]] 
X_train:[[0 1]
 [4 5]
 [6 7]
 [8 9]] 
X_test: [[2 3]] 
X_train:[[0 1]
 [2 3]
 [6 7]
 [8 9]] 
X_test: [[4 5]] 
X_train:[[0 1]
 [2 3]
 [4 5]
 [8 9]] 
X_test: [[6 7]] 
X_train:[[0 1]
 [2 3]
 [4 5]
 [6 7]] 
X_test: [[8 9]] 

三、自助法(bootstrapping)