Sklearn 多标签 ML:值错误:标签二值化不支持多输出目标数据

Question

我正在构建一个程序，该程序为文本描述分配多个标签/标签。我正在使用OneVsRestClassifier来标记我的文本描述。xTrain，xTest和yTrain都是'numpy.ndarray'的。考虑到我以正确的方式拆分了训练和测试数据，这似乎很奇怪。下面是我的代码：

xTrain, xTest, yTrain, yTest = train_test_split(x, y, test_size=0.2)
nb_clf = MultinomialNB()
sgd = SGDClassifier()
lr = LogisticRegression()
mn = MultinomialNB()
print("xTrain.shape = " + str(xTrain.shape))
print("xTest.shape = " + str(xTest.shape))
print("yTrain.shape = " + str(yTrain.shape))
print("yTest.shape = " + str(yTest.shape))
print("type(xTrain) = " + str(type(xTrain)))
print("type(xTest) = " + str(type(xTest)))
xTrain = csr_matrix(xTrain).toarray()
xTest = csr_matrix(xTest).toarray()
yTrain = csr_matrix(yTrain).toarray()
print("type(xTrain) = " + str(type(xTrain)))
for classifier in [nb_clf, sgd, lr, mn]:
    clf = OneVsRestClassifier(classifier)
    clf.fit(xTrain.astype("U"), yTrain.astype("U"))
    y_pred = clf.predict(xTest)
    print("ny_pred:")
    print(y_pred)

x 输出：

  (1466, 1292)  0.13531037414782607
  (1466, 1238)  0.21029405543816293
  (1466, 988)   0.04688335706505732
  ...
  ...

y 输出：

[[0 0 0 ... 1 0 0]
 [0 0 0 ... 0 0 0]
 [0 0 0 ... 1 0 0]
 ...
 [0 0 0 ... 0 0 0]
 [0 0 0 ... 0 0 0]
 [0 0 0 ... 0 0 0]]

打印语句输出：

xTrain.shape = (1173, 13817)
xTest.shape = (294, 13817)
yTrain.shape = (1173, 28)
yTest.shape = (294, 28)
type(xTrain) = <class 'scipy.sparse.csr.csr_matrix'>
type(xTest) = <class 'scipy.sparse.csr.csr_matrix'>
type(xTrain) = <class 'numpy.ndarray'>
type(xTest) = <class 'numpy.ndarray'>
type(yTrain) = <class 'numpy.ndarray'>

错误(在 clf.fit 行(：

值错误：标签不支持多输出目标数据 二值化

Answer 1

请首先澄清程序中的特征维度和样本大小。对于目标特征(y(，标签不应该是独热编码的。例如，它应该是 [3] 而不是 [0 0 0 1]。