Python, SelectPercentile.fit() returns 异常矩阵和错误，但是为什么呢？

Question

我是机器学习的新手，正在构建一个文本分类系统。我的文本是疾病的症状，标签是疾病的名称。我已经对所有文本进行了预处理并计算了 tf-idf values.As 我的训练集和测试集很大，所以我没有提到任何人需要的内容 here.If，我也可以包括 later.Here是labels_train，训练集和测试集的代码和输出tf-idf矩阵：

labels_train = ['Heart Attack', 'Heart Failure', 'Asthma', 'High Blood Pressure', 'Dengue Fever', 'Lung Cancer']

代码：

from sklearn import cross_validation
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.feature_selection import SelectPercentile, f_classif

vectorizer=TfidfVectorizer(sublinear_tf=True,max_df=0.5,stop_words='english')
features_train_transformed = vectorizer.fit_transform(features_train).toarray()
print "features_train: ",features_train_transformed
features_test_transformed  = vectorizer.transform(features_test)
print "features_test: ",features_test_transformed

输出：

features_train:  [[ 0.          0.09420201  0.         ...,  0.          0.          0.06521722]
 [ 0.          0.          0.         ...,  0.          0.          0.        ]
 [ 0.          0.          0.         ...,  0.          0.          0.        ]
 [ 0.          0.          0.         ...,  0.185467    0.08837602
   0.06118382]
 [ 0.11319809  0.          0.11319809 ...,  0.          0.          0.07836844]
 [ 0.          0.          0.         ...,  0.          0.          0.        ]]


    features_test:  [[ 0.          0.          0.          0.          0.          0.          0.
    0.          0.          0.          0.          0.          0.          0.
    0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.50792975  0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.16756124  0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.24203125  0.          0.
   0.          0.          0.          0.          0.          0.19846908
   0.          0.          0.40979452  0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.
   0.19846908  0.          0.          0.          0.          0.
   0.19846908  0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.24203125
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.19846908  0.          0.          0.
   0.          0.          0.          0.          0.          0.16756124
   0.          0.          0.          0.          0.          0.
   0.24203125  0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.24203125  0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.19846908  0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.16756124  0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.24203125  0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.          0.          0.          0.          0.          0.          0.
   0.        ]]

问题是，当我想要 select 特征时，我得到了一个零矩阵。

代码：

selector = SelectPercentile(f_classif, percentile=5)
selector.fit(features_train_transformed, labels_train)

输出：

0 [ 0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.
  0.  0.]
SelectPercentile(percentile=5,
         score_func=<function f_classif at 0x00000000053F6278>)

我不明白为什么我会收到这个。我怎样才能摆脱它或可能的错误在哪里？

Answer 1

我找到了解决方案

这里唯一的问题是只有几个样本，不足以计算出方差分析（方差分析），因为方差分析不能处理零方差特征。我应该只增加火车数据集的数量。我从这些链接中找到了详细信息： https://www.quora.com/How-is-ANOVA-used-for-feature-selection

https://github.com/scikit-learn/scikit-learn/issues/462

Python, SelectPercentile.fit() returns 异常矩阵和错误，但是为什么呢？

In Python, SelectPercentile.fit() returns unusual matrix and error,but Why?

feature-selection

python-2.7