最初,我从 .csv
中读取数据文件,但在这里我从列表构建数据框,以便可以重现问题。目的是使用 LogisticRegressionCV
训练具有交叉验证的逻辑回归模型。 .
indeps = ['M', 'F', 'M', 'F', 'M', 'M', 'F', 'M', 'M', 'F', 'F', 'F', 'F', 'F', 'M', 'F', 'F', 'F', 'F', 'F', 'M', 'F', 'F', 'M', 'M', 'F', 'F', 'F', 'M', 'F', 'F', 'F', 'M', 'F', 'M', 'F', 'F', 'F', 'M', 'M', 'M', 'F', 'M', 'M', 'M', 'F', 'M', 'M', 'F', 'F']
dep = [1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
data = [indeps, dep]
cols = ['state', 'cat_bins']
data_dict = dict((x[0], x[1]) for x in zip(cols, data))
df = pd.DataFrame.from_dict(data_dict)
df.tail()
cat_bins state
45 0.0 F
46 0.0 M
47 0.0 M
48 0.0 F
49 0.0 F
'''Use Pandas' to encode independent variables. Notice that
we are returning a sparse dataframe '''
def heat_it2(dataframe, lst_of_columns):
dataframe_hot = pd.get_dummies(dataframe,
prefix = lst_of_columns,
columns = lst_of_columns, sparse=True,)
return dataframe_hot
train_set_hot = heat_it2(df, ['state'])
train_set_hot.head(2)
cat_bins state_F state_M
0 1.0 0 1
1 1.0 1 0
'''Use the dataframe to set up the prospective inputs to the model as numpy arrays'''
indeps_hot = ['state_F', 'state_M']
X = train_set_hot[indeps_hot].values
y = train_set_hot['cat_bins'].values
print 'X-type:', X.shape, type(X)
print 'y-type:', y.shape, type(y)
print 'X has shape, is an array and has length:\n', hasattr(X, 'shape'), hasattr(X, '__array__'), hasattr(X, '__len__')
print 'yhas shape, is an array and has length:\n', hasattr(y, 'shape'), hasattr(y, '__array__'), hasattr(y, '__len__')
print 'X does have attribute fit:\n',hasattr(X, 'fit')
print 'y does have attribute fit:\n',hasattr(y, 'fit')
X-type: (50, 2) <type 'numpy.ndarray'>
y-type: (50,) <type 'numpy.ndarray'>
X has shape, is an array and has length:
True True True
yhas shape, is an array and has length:
True True True
X does have attribute fit:
False
y does have attribute fit:
False
因此,回归器的输入似乎具有 .fit
的必要属性方法。它们是 具有正确形状的 numpy 数组。 X
是维度为 [n_samples, n_features]
的数组, 和 y
是一个形状为 [n_samples,]
的向量这是文档:
fit(X, y, sample_weight=None)[source]
Fit the model according to the given training data. Parameters: X : {array-like, sparse matrix}, shape (n_samples, n_features) Training vector, where n_samples is the number of samples and n_features is the number of features. y : array-like, shape (n_samples,) Target vector relative to X.
....
现在我们尝试拟合回归量:
logmodel = LogisticRegressionCV(Cs =1, dual=False , scoring = accuracy_score, penalty = 'l2')
logmodel.fit(X, y)
...
TypeError: Expected sequence or array-like, got estimator LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,
intercept_scaling=1, max_iter=100, multi_class='ovr', n_jobs=1,
penalty='l2', random_state=None, solver='liblinear', tol=0.0001,
verbose=0, warm_start=False)
错误消息的来源似乎在 scikits 的 validation.py 模块中,here .
引发此错误消息的唯一代码部分是以下函数片段:
def _num_samples(x):
"""Return number of samples in array-like x."""
if hasattr(x, 'fit'):
# Don't get num_samples from an ensembles length!
raise TypeError('Expected sequence or array-like, got '
'estimator %s' % x)
etc.
问题:由于我们用来拟合模型的参数(X
和 y
)不具有“fit”属性,为什么会出现此错误消息
在带有 scikit-learn 18.01-3 和 pandas 0.19.2-2 的 Canopy 1.7.4.3348(64 位)上使用 python 2.7
谢谢你的帮助:)
最佳答案
问题似乎出在 scoring
参数中。您已通过 accuracy_score
。 accuracy_score
的签名是 accuracy_score(y_true, y_pred[, ...])
。但是在模块logistic.py
if isinstance(scoring, six.string_types):
scoring = SCORERS[scoring]
for w in coefs:
// Other code
if scoring is None:
scores.append(log_reg.score(X_test, y_test))
else:
scores.append(scoring(log_reg, X_test, y_test))
因为你已经通过了 accuracy_score
,所以它不符合上面的第一行。
并且 scores.append(scoring(log_reg, X_test, y_test))
用于对估计器进行评分。但正如我上面所说,这里的参数与 accuracy_score
所需的参数不匹配。因此错误。
解决方法:使用 make_scorer LogisticRegressionCV 中的 (accuracy_score) 用于评分或简单地传递字符串 'accuracy'
logmodel = LogisticRegressionCV(Cs =1, dual=False ,
scoring = make_scorer(accuracy_score),
penalty = 'l2')
OR
logmodel = LogisticRegressionCV(Cs =1, dual=False ,
scoring = 'accuracy',
penalty = 'l2')
注意:
这可能是 logistic.py
模块的一部分或 LogisticRegressionCV 文档中的错误,他们应该已经阐明了评分函数的签名。
关于python - Sklearn LogisticRegressionCV 的类似数组的输入,我们在Stack Overflow上找到一个类似的问题: https://stackoverflow.com/questions/42151921/