如何在sklearn管道中通过功能消除选择功能名称?

时间:2016-04-14 20:34:57

标签: python machine-learning scikit-learn

我在sklearn管道中使用递归功能消除,管道看起来像这样:

from sklearn.pipeline import FeatureUnion, Pipeline
from sklearn import feature_selection
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.svm import LinearSVC

X = ['I am a sentence', 'an example']
Y = [1, 2]
X_dev = ['another sentence']

# classifier
LinearSVC1 = LinearSVC(tol=1e-4,  C = 0.10000000000000001)
f5 = feature_selection.RFE(estimator=LinearSVC1, n_features_to_select=500, step=1)

pipeline = Pipeline([
    ('features', FeatureUnion([
       ('tfidf', TfidfVectorizer(ngram_range=(1, 3), max_features= 4000)), 
       ('custom_features', CustomFeatures())])),
    ('rfe_feature_selection', f5),
    ('clf', LinearSVC1),
    ])

pipeline.fit(X, Y)
y_pred = pipeline.predict(X_dev)

如何获取RFE选择的功能的功能名称? RFE应该选择最好的500个功能,但我真的需要看看已经选择了哪些功能。

编辑:

我有一个复杂的管道,它包含多个管道和特征联合,百分位特征选择以及最后的递归特征消除:

fs = feature_selection.SelectPercentile(feature_selection.chi2, percentile=90)
fs_vect = feature_selection.SelectPercentile(feature_selection.chi2, percentile=80)
f5 = feature_selection.RFE(estimator=svc, n_features_to_select=600, step=3)

countVecWord = TfidfVectorizer(ngram_range=(1, 3), max_features=2000, analyzer=u'word', sublinear_tf=True, use_idf = True, min_df=2, max_df=0.85, lowercase = True)
countVecWord_tags = TfidfVectorizer(ngram_range=(1, 4), max_features= 1000, analyzer=u'word', min_df=2, max_df=0.85, sublinear_tf=True, use_idf = True, lowercase = False)

pipeline = Pipeline([
        ('union', FeatureUnion(
                transformer_list=[

                ('vectorized_pipeline', Pipeline([
                    ('union_vectorizer', FeatureUnion([

                        ('stem_text', Pipeline([
                            ('selector', ItemSelector(key='stem_text')),
                            ('stem_tfidf', countVecWord)
                        ])),

                        ('pos_text', Pipeline([
                            ('selector', ItemSelector(key='pos_text')),
                            ('pos_tfidf', countVecWord_tags)
                        ])),

                    ])),
                        ('percentile_feature_selection', fs_vect)
                    ])),


                ('custom_pipeline', Pipeline([
                    ('custom_features', FeatureUnion([

                        ('pos_cluster', Pipeline([
                            ('selector', ItemSelector(key='pos_text')),
                            ('pos_cluster_inner', pos_cluster)
                        ])),

                        ('stylistic_features', Pipeline([
                            ('selector', ItemSelector(key='raw_text')),
                            ('stylistic_features_inner', stylistic_features)
                        ])),


                    ])),
                        ('percentile_feature_selection', fs),
                        ('inner_scale', inner_scaler)
                ])),

                ],

                # weight components in FeatureUnion
                # n_jobs=6,

                transformer_weights={
                    'vectorized_pipeline': 0.8,  # 0.8,
                    'custom_pipeline': 1.0  # 1.0
                },
        )),

        ('rfe_feature_selection', f5),
        ('clf', classifier),
        ])

我会尝试解释这些步骤。第一个Pipeline由向量化器组成,称为" vectorized_pipeline"所有这些都有一个函数" get_feature_names"。第二个Pipeline由我自己的特性组成,我已经使用fit,transform和get_feature_names函数实现了它们。当我使用@Kevin的建议时,我收到一个错误,即' union' (这是我在管道中的顶部元素的名称)没有get_feature_names函数:

support = pipeline.named_steps['rfe_feature_selection'].support_
feature_names = pipeline.named_steps['union'].get_feature_names()
print np.array(feature_names)[support]

此外,当我尝试从各个FeatureUnion获取功能名称时,如下所示:

support = pipeline.named_steps['rfe_feature_selection'].support_
feature_names = pipeline_age.named_steps['union_vectorizer'].get_feature_names()
print np.array(feature_names)[support]

我收到一个关键错误:

feature_names = pipeline.named_steps['union_vectorizer'].get_feature_names()
KeyError: 'union_vectorizer'

1 个答案:

答案 0 :(得分:8)

您可以使用属性named_steps访问Pipeline的每个步骤,这是iris数据集上的一个示例,它只选择2个功能,但解决方案会缩放。

from sklearn import datasets
from sklearn import feature_selection
from sklearn.svm import LinearSVC

iris = datasets.load_iris()
X = iris.data
y = iris.target

# classifier
LinearSVC1 = LinearSVC(tol=1e-4,  C = 0.10000000000000001)
f5 = feature_selection.RFE(estimator=LinearSVC1, n_features_to_select=2, step=1)

pipeline = Pipeline([
    ('rfe_feature_selection', f5),
    ('clf', LinearSVC1)
    ])

pipeline.fit(X, y)

使用named_steps,您可以访问管道中变换对象的属性和方法。 RFE属性support_(或方法get_support())将返回所选要素的布尔掩码:

support = pipeline.named_steps['rfe_feature_selection'].support_

现在support是一个数组,您可以使用它来有效地提取所选要素(列)的名称。确保您的功能名称位于numpy array,而不是python列表。

import numpy as np
feature_names = np.array(iris.feature_names) # transformed list to array

feature_names[support]

array(['sepal width (cm)', 'petal width (cm)'], 
      dtype='|S17')

修改

根据我上面的评论,以下是删除了CustomFeautures()函数的示例:

from sklearn.pipeline import FeatureUnion, Pipeline
from sklearn import feature_selection
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.svm import LinearSVC
import numpy as np

X = ['I am a sentence', 'an example']
Y = [1, 2]
X_dev = ['another sentence']

# classifier
LinearSVC1 = LinearSVC(tol=1e-4,  C = 0.10000000000000001)
f5 = feature_selection.RFE(estimator=LinearSVC1, n_features_to_select=500, step=1)

pipeline = Pipeline([
    ('features', FeatureUnion([
       ('tfidf', TfidfVectorizer(ngram_range=(1, 3), max_features= 4000))])), 
    ('rfe_feature_selection', f5),
    ('clf', LinearSVC1),
    ])

pipeline.fit(X, Y)
y_pred = pipeline.predict(X_dev)

support = pipeline.named_steps['rfe_feature_selection'].support_
feature_names = pipeline.named_steps['features'].get_feature_names()
np.array(feature_names)[support]