Python Scikit-learn的多类分类出错

Question

我试图在Scikit-learn中可用的分类器之间实现一点比较。根据这个页面，除了svm之外，所有的分类器都应该工作。

该操作的实现方法如下:

clf['bayes'] = OneVsRestClassifier(MultinomialNB(
clf['lda'] = OneVsRestClassifier(LDA())
clf['decision tree'] = OneVsRestClassifier(DecisionTreeClassifier())
clf['rdc'] = OneVsRestClassifier(RandomForestClassifier())

y_supposes = {}
precision = {}
for classifier in clf:
    clf[classifier].fit(x_train, y_train)
    y_supposes[classifier] = clf[classifier].predict(x_test)
    precision[classifier] = calcul_precision(y_supposes[classifier], y_test)

问题是，唯一有效的分类器是bayes分类器。

当我尝试调用classifier['rdc'].fit(x_train, y_train)时，另一个给我这个错误:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "C:Python27libsite-packagessklearnmulticlass.py", line 201, in fit
    n_jobs=self.n_jobs)
  File "C:Python27libsite-packagessklearnmulticlass.py", line 92, in fit_ov
r
    for i in range(Y.shape[1]))
  File "C:Python27libsite-packagessklearnexternalsjoblibparallel.py", lin
e 517, in __call__
    self.dispatch(function, args, kwargs)
  File "C:Python27libsite-packagessklearnexternalsjoblibparallel.py", lin
e 312, in dispatch
    job = ImmediateApply(func, args, kwargs)
  File "C:Python27libsite-packagessklearnexternalsjoblibparallel.py", lin
e 136, in __init__
    self.results = func(*args, **kwargs)
  File "C:Python27libsite-packagessklearnmulticlass.py", line 61, in _fit_b
inary
    estimator.fit(X, y)
  File "C:Python27libsite-packagessklearnensembleforest.py", line 257, in
fit
    check_ccontiguous=True)
  File "C:Python27libsite-packagessklearnutilsvalidation.py", line 220, in
 check_arrays
    raise TypeError('A sparse matrix was passed, but dense '
TypeError: A sparse matrix was passed, but dense data is required. Use X.toarray
() to convert to a dense numpy array.

我想补充的是，clf['rdc'].fit(x_train.toarray, y_train)(如错误信息所示)也给了我一个错误。

你能帮我找到我跳过的步骤吗?

编辑:新进展

我认为问题可能来自x_train的类型。我的计算方法如下:

x = [{f1 : a, ... fn : jo}, ..., {f3 : 5}]
y_train = [('label1', ), ..., ('labelZ', 'label72')]
x_train = DictVectorizer.fit_transform(x)
type(x_train) ==  <class 'scipy.sparse.csr.csr_matrix'>

我也尝试了这种方法:MultinomialNB.fit(np.array(x), np.array(y))，它给了我一个新的错误信息:

Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "C:Python27libsite-packagessklearnnaive_bayes.py", line 308, in fit
X = X.astype(np.float)
TypeError: float() argument must be a string or a number

Answer 1

错误消息非常清楚地表明，您正在将稀疏矩阵传递给不支持稀疏矩阵的估计器。在您测试的四个分类器中，只有MultinomialNB支持稀疏矩阵输入。对于决策树和随机森林，稀疏矩阵支持正在进行中。

至于np.array(x)，它并不像你想象的那样。要将稀疏矩阵转换为密集数组，请使用x.toarray()，或者将sparse=False传递给DictVectorizer构造函数。