# -*- coding: utf-8 -*-
"""
Created on Mon Aug 21 10:57:29 2017
@author: 飄的心
"""
#過(guò)濾式特征選擇
#根據(jù)方差進(jìn)行選擇，方差越小，代表該屬性識(shí)別能力很差，可以剔除
from sklearn.feature_selection import VarianceThreshold
x=[[100,1,2,3],
  [100,4,5,6],
  [100,7,8,9],
  [101,11,12,13]]
selector=VarianceThreshold(1) #方差閾值值，
selector.fit(x)
selector.variances_ #展現(xiàn)屬性的方差
selector.transform(x)#進(jìn)行特征選擇
selector.get_support(True) #選擇結(jié)果后，特征之前的索引
selector.inverse_transform(selector.transform(x)) #將特征選擇后的結(jié)果還原成原始數(shù)據(jù)
                         #被剔除掉的數(shù)據(jù)，顯示為0
#單變量特征選擇
from sklearn.feature_selection import SelectKBest,f_classif
x=[[1,2,3,4,5],
  [5,4,3,2,1],
  [3,3,3,3,3],
  [1,1,1,1,1]]
y=[0,1,0,1]
selector=SelectKBest(score_func=f_classif,k=3)#選擇3個(gè)特征，指標(biāo)使用的是方差分析F值
selector.fit(x,y)
selector.scores_ #每一個(gè)特征的得分
selector.pvalues_
selector.get_support(True) #如果為true，則返回被選出的特征下標(biāo)，如果選擇False，則
              #返回的是一個(gè)布爾值組成的數(shù)組，該數(shù)組只是那些特征被選擇
selector.transform(x)
#包裹時(shí)特征選擇
from sklearn.feature_selection import RFE
from sklearn.svm import LinearSVC #選擇svm作為評(píng)定算法
from sklearn.datasets import load_iris #加載數(shù)據(jù)集
iris=load_iris()
x=iris.data
y=iris.target
estimator=LinearSVC()
selector=RFE(estimator=estimator,n_features_to_select=2) #選擇2個(gè)特征
selector.fit(x,y)
selector.n_features_  #給出被選出的特征的數(shù)量
selector.support_   #給出了被選擇特征的mask
selector.ranking_   #特征排名，被選出特征的排名為1
#注意：特征提取對(duì)于預(yù)測(cè)性能的提升沒(méi)有必然的聯(lián)系，接下來(lái)進(jìn)行比較；
from sklearn.feature_selection import RFE
from sklearn.svm import LinearSVC
from sklearn import cross_validation
from sklearn.datasets import load_iris
#加載數(shù)據(jù)
iris=load_iris()
X=iris.data
y=iris.target
#特征提取
estimator=LinearSVC()
selector=RFE(estimator=estimator,n_features_to_select=2)
X_t=selector.fit_transform(X,y)
#切分測(cè)試集與驗(yàn)證集
x_train,x_test,y_train,y_test=cross_validation.train_test_split(X,y,
                  test_size=0.25,random_state=0,stratify=y)
x_train_t,x_test_t,y_train_t,y_test_t=cross_validation.train_test_split(X_t,y,
                  test_size=0.25,random_state=0,stratify=y)
clf=LinearSVC()
clf_t=LinearSVC()
clf.fit(x_train,y_train)
clf_t.fit(x_train_t,y_train_t)
print('origin dataset test score:',clf.score(x_test,y_test))
#origin dataset test score: 0.973684210526
print('selected Dataset:test score:',clf_t.score(x_test_t,y_test_t))
#selected Dataset:test score: 0.947368421053
import numpy as np
from sklearn.feature_selection import RFECV
from sklearn.svm import LinearSVC
from sklearn.datasets import load_iris
iris=load_iris()
x=iris.data
y=iris.target
estimator=LinearSVC()
selector=RFECV(estimator=estimator,cv=3)
selector.fit(x,y)
selector.n_features_
selector.support_
selector.ranking_
selector.grid_scores_
#嵌入式特征選擇
import numpy as np
from sklearn.feature_selection import SelectFromModel
from sklearn.svm import LinearSVC
from sklearn.datasets import load_digits
digits=load_digits()
x=digits.data
y=digits.target
estimator=LinearSVC(penalty='l1',dual=False)
selector=SelectFromModel(estimator=estimator,threshold='mean')
selector.fit(x,y)
selector.transform(x)
selector.threshold_
selector.get_support(indices=True)
#scikitlearn提供了Pipeline來(lái)講多個(gè)學(xué)習(xí)器組成流水線，通常流水線的形式為：將數(shù)據(jù)標(biāo)準(zhǔn)化，
#--》特征提取的學(xué)習(xí)器————》執(zhí)行預(yù)測(cè)的學(xué)習(xí)器，除了最后一個(gè)學(xué)習(xí)器之后，
#前面的所有學(xué)習(xí)器必須提供transform方法，該方法用于數(shù)據(jù)轉(zhuǎn)化（如歸一化、正則化、
#以及特征提取
#學(xué)習(xí)器流水線（pipeline）
from sklearn.svm import LinearSVC
from sklearn.datasets import load_digits
from sklearn import cross_validation
from sklearn.linear_model import LogisticRegression
from sklearn.pipeline import Pipeline
def test_Pipeline(data):
  x_train,x_test,y_train,y_test=data
  steps=[('linear_svm',LinearSVC(C=1,penalty='l1',dual=False)),
      ('logisticregression',LogisticRegression(C=1))]
  pipeline=Pipeline(steps)
  pipeline.fit(x_train,y_train)
  print('named steps',pipeline.named_steps)
  print('pipeline score',pipeline.score(x_test,y_test))
if __name__=='__main__':
  data=load_digits()
  x=data.data
  y=data.target
  test_Pipeline(cross_validation.train_test_split(x,y,test_size=0.25,
                  random_state=0,stratify=y))