特征选择时出错答案

【问题标题】：Error during feature selection特征选择时出错
【发布时间】：2016-04-01 09:43:55
【问题描述】：

我正在尝试为多标签分类进行特征选择。我提取了模型将被训练到 X 中的特征。模型测试是在同一个 X 上完成的。我正在使用 Pipeline 并选择最好的 100 个特征-

#arrFinal contains all the features and the labels. Last 16 columns are labels and features are from 1 to 521. 17th column from the last is not taken
X=np.array(arrFinal[:,1:-17])
Xtest=np.array(X)
Y=np.array(arrFinal[:,522:]).astype(int)
clf = Pipeline([('chi2', SelectKBest(chi2, k=100)),('rbf',SVC())])
clf = OneVsRestClassifier(clf)
clf.fit(X, Y)
ans=clf.predict(X_test)

但我收到以下错误-

Traceback (most recent call last):
  File "C:\Users\50004182\Documents\\callee.py", line 10, in <module
>
    combine.combine_main(dict_ids,inv_dict_ids,noOfIDs)
  File "C:\Users\50004182\Documents\combine.py", line 201, in combi
ne_main
    clf.fit(X, Y)
  File "C:\Python34\lib\site-packages\sklearn\multiclass.py", line 287, in fit
    for i, column in enumerate(columns))
  File "C:\Python34\lib\site-packages\sklearn\externals\joblib\parallel.py", lin
e 804, in __call__
    while self.dispatch_one_batch(iterator):
  File "C:\Python34\lib\site-packages\sklearn\externals\joblib\parallel.py", lin
e 662, in dispatch_one_batch
    self._dispatch(tasks)
  File "C:\Python34\lib\site-packages\sklearn\externals\joblib\parallel.py", lin
e 570, in _dispatch
    job = ImmediateComputeBatch(batch)
  File "C:\Python34\lib\site-packages\sklearn\externals\joblib\parallel.py", lin
e 183, in __init__
    self.results = batch()
  File "C:\Python34\lib\site-packages\sklearn\externals\joblib\parallel.py", lin
e 72, in __call__
    return [func(*args, **kwargs) for func, args, kwargs in self.items]
  File "C:\Python34\lib\site-packages\sklearn\externals\joblib\parallel.py", lin
e 72, in <listcomp>
    return [func(*args, **kwargs) for func, args, kwargs in self.items]
  File "C:\Python34\lib\site-packages\sklearn\multiclass.py", line 74, in _fit_b
inary
    estimator.fit(X, y)
  File "C:\Python34\lib\site-packages\sklearn\pipeline.py", line 164, in fit
    Xt, fit_params = self._pre_transform(X, y, **fit_params)
  File "C:\Python34\lib\site-packages\sklearn\pipeline.py", line 145, in _pre_tr
ansform
    Xt = transform.fit_transform(Xt, y, **fit_params_steps[name])
  File "C:\Python34\lib\site-packages\sklearn\base.py", line 458, in fit_transfo
rm
    return self.fit(X, y, **fit_params).transform(X)
  File "C:\Python34\lib\site-packages\sklearn\feature_selection\univariate_selec
tion.py", line 331, in fit
    self.scores_, self.pvalues_ = self.score_func(X, y)
  File "C:\Python34\lib\site-packages\sklearn\feature_selection\univariate_selec
tion.py", line 213, in chi2
    if np.any((X.data if issparse(X) else X) < 0):
TypeError: unorderable types: numpy.ndarray() < int()

【问题讨论】：

我无法完全描绘arrFinal 中的数据。您能否发布一个 sn-p 以生成一些虚拟数据或更全面地描述它的内容？
另外，看起来您可能在您的 arrFinal 数据中的某处有一个 ndarray，您希望在该数据中有一个 int。您是否检查了所有列的dtype？
你能检查X.dtype的值吗？看起来这是最有可能的罪魁祸首。
我以前没见过这种类型。尝试使用 X=np.array(arrFinal[:,1:-17]).astype(np.float64) 构建 X
看起来chi2 不喜欢只包含0s 的列。尝试通过X = VarianceThreshold().fit_transform(X) 传递它，或者只是将其添加到您的管道中。

标签： python machine-learning scikit-learn feature-selection

【解决方案1】：

因此，在与@JamieBull 和@Joker 进行上述cmets 的调试会话之后。我们提出的解决方案是：

确保类型正确（原字符串）

X=np.array(arrFinal[:,1:-17]).astype(np.float64)
Xtest=np.array(X)
Y=np.array(arrFinal[:,522:]).astype(int)

首先使用VarianceThreshold 删除chi2 之前的常量(0) 列。

clf = Pipeline([
      ('vt', VarianceThreshold()),
      ('chi2', SelectKBest(chi2, k=100)),
      ('rbf',SVC())
])
clf = OneVsRestClassifier(clf)
clf.fit(X, Y)
ans=clf.predict(X_test)

【讨论】：