sklearn之Multiclass 估计器

sklearn.multiclass 可以处理多类别 (multi-class) 的多标签 (multi-label) 的分类问题。

多类别分类

手写数字有 0-9 十类，但手头上只有两分类估计器 (比如像支撑向量机) 怎么用呢？我们可以采取下面三种常见策略：

• 一对一 (One vs One, OvO)：一个分类器用来处理数字 0 和数字 1，一个用来处理数字 0 和数字 2，一个用来处理数字 1 和 2，以此类推。N 个类需要 N(N-1)/2 个分类器。

• 一对其他 (One vs All, OvA)：训练 10 个二分类器，每一个对应一个数字，第一个分类 1 和「非1」，第二个分类 2 和「非2」，以此类推。N 个类需要 N 个分类器。

OneVsOneClassifier：

from sklearn.datasets import load_digits
from sklearn.model_selection import train_test_split
import matplotlib.pyplot as plt
from sklearn.multiclass import OneVsOneClassifier
from sklearn.linear_model import LogisticRegression
from sklearn import metrics#metrics 来计算各种性能指标

digits = load_digits()

X_train, X_test, y_train, y_test= train_test_split( digits['data'], digits['target'], test_size=0.2 )
print( 'The size of X_train is ', X_train.shape )
print( 'The size of y_train is ', y_train.shape )
print( 'The size of X_test is ', X_test.shape )
print( 'The size of y_test is ', y_test.shape )

fig,axes = plt.subplots(10,10,figsize=(8,8))
fig.subplots_adjust(hspace=0.1,wspace=0.1)

for i,ax in enumerate(axes.flat):
    ax.imshow(X_train[i,:].reshape(8,8),cmap="binary",interpolation="nearest")
    ax.text(0.05,0.05,str(y_train[i]),transform=ax.transAxes,color="blue")
    ax.set_xticks([])
    ax.set_yticks([])

plt.show()



ovo_lr=OneVsOneClassifier(LogisticRegression(solver="lbfgs",max_iter=200))#创建一个一对一的多分类器
ovo_lr.fit(X_train,y_train) #开始分类

print(len(ovo_lr.estimators_))#查看分类器的数量
print("train_OVO LR",metrics.accuracy_score(y_train,ovo_lr.predict(X_train)))
print("test_ovo LR",metrics.accuracy_score(y_test,ovo_lr.predict(X_test)))

查看结果：

F:\开发工具\pythonProject\tools\venv\Scripts\python.exe F:/开发工具/pythonProject/tools/python的sklear学习/sklearn01.py
The size of X_train is  (1437, 64)
The size of y_train is  (1437,)
The size of X_test is  (360, 64)
The size of y_test is  (360,)
45
train_OVO LR 1.0
test_ovo LR 0.9833333333333333

Process finished with exit code 0

OneVsRestClassifier：

from sklearn.datasets import load_digits
from sklearn.model_selection import train_test_split
from sklearn.multiclass import OneVsRestClassifier
from sklearn import metrics#metrics 来计算各种性能指标
from sklearn.linear_model import LogisticRegression

digits = load_digits()
X_train, X_test, y_train, y_test= train_test_split( digits['data'], digits['target'], test_size=0.2 )

ova_lr=OneVsRestClassifier(LogisticRegression(solver="lbfgs",max_iter=800))

ova_lr.fit(X_train,y_train)



print( len(ova_lr.estimators_) )	#查看分类器的数量
print("train_ova_lr",metrics.accuracy_score(y_train,ova_lr.predict(X_train)))
print("test_ova_lr",metrics.accuracy_score(y_test,ova_lr.predict(X_test)))

测试结果：

F:\开发工具\pythonProject\tools\venv\Scripts\python.exe F:/开发工具/pythonProject/tools/python的sklear学习/sklearn02.py
10
train_ova_lr 0.9972164231036882
test_ova_lr 0.9527777777777777

Process finished with exit code 0

多标签分类：

我们特意为每个数字设计了多标签：

• 标签 1 - 奇数、偶数

• 标签 2 - 小于等于 4，大于 4

 

from sklearn.datasets import load_digits
from sklearn.model_selection import train_test_split
from sklearn.multiclass import OneVsRestClassifier
from sklearn import metrics# metrics 来计算各种性能指标
from sklearn.linear_model import LogisticRegression
import numpy as np

digits = load_digits()
X_train, X_test, y_train, y_test= train_test_split( digits['data'], digits['target'], test_size=0.2 )
y_train_multilabel 	= np.c_[ y_train%2==0, y_train<=4 ]
ova_lr =OneVsRestClassifier(LogisticRegression(solver="lbfgs" ,max_iter=800))

ova_lr.fit(X_train ,y_train_multilabel)


print(y_train_multilabel)
print( len(ova_lr.estimators_) )  # 查看分类器的数量
print( y_test[:1] )
print( ova_lr.predict(X_test[:1,:]) )

 测试结果：

F:\开发工具\pythonProject\tools\venv\Scripts\python.exe F:/开发工具/pythonProject/tools/python的sklear学习/sklearn03.py
[[False False]
 [False False]
 [ True False]
 ...
 [ True False]
 [False  True]
 [False  True]]
2
[6]
[[1 0]]

Process finished with exit code 0