ML之sklearn:sklearn.linear_mode中的LogisticRegression函数的简介、使用方法之详细攻略(一)

sklearn.linear_mode中的LogisticRegression函数的简介、使用方法



class LogisticRegression Found at: sklearn.linear_model._logisticclass LogisticRegression(BaseEstimator, LinearClassifierMixin,  SparseCoefMixin):

   """

   Logistic Regression (aka logit, MaxEnt) classifier.

   In the multiclass case, the training algorithm uses the one-vs-rest (OvR) scheme if the 'multi_class' option is set to 'ovr', and uses the cross-entropy loss if the 'multi_class' option is set to 'multinomial'. (Currently the 'multinomial' option is supported only by the 'lbfgs', 'sag', 'saga' and 'newton-cg' solvers.)

   

   This class implements regularized logistic regression using the 'liblinear' library, 'newton-cg', 'sag', 'saga' and 'lbfgs' solvers. **Note that regularization is applied by default**. It can handle both dense and sparse input. Use C-ordered arrays or CSR matrices containing 64-bit floats for optimal performance; any other input format will be converted (and copied).

   

   The 'newton-cg', 'sag', and 'lbfgs' solvers support only L2 regularization with primal formulation, or no regularization. The 'liblinear' solver supports both L1 and L2 regularization, with a dual formulation only for the L2 penalty. The Elastic-Net regularization is only supported by the 'saga' solver.

   

   Read more in the :ref:`User Guide <logistic_regression>`.

逻辑回归(又名logit, MaxEnt)分类器。

在多类情况下,如果“multi_class”选项设置为“OvR”,训练算法使用one vs-rest (OvR)方案,如果“multi_class”选项设置为“多项”,训练算法使用交叉熵损失。(目前,“多项”选项仅由“lbfgs”、“sag”、“saga”和“newton-cg”求解器支持。)


这个类使用“liblinear”库、“newton-cg”、“sag”、“saga”和“lbfgs”求解器实现正则逻辑回归。**注意正则化是在默认情况下应用的**。它可以处理稠密和稀疏输入。使用C-ordered数组或包含64位浮点数的CSR矩阵,以获得最佳性能;任何其他输入格式都将被转换(和复制)。


“newton-cg”、“sag”和“lbfgs”求解器只支持使用原始公式的L2正则化,或者不支持正则化。“liblinear”求解器支持L1和L2正则化,只有L2惩罚的对偶公式。弹性网正则化仅由“saga”求解器支持。


详见:ref: ' User Guide <logistic_regression> '。</logistic_regression>

 Parameters

   ----------

   penalty : {'l1', 'l2', 'elasticnet', 'none'}, default='l2'

   Used to specify the norm used in the penalization. The 'newton-cg', 'sag' and 'lbfgs' solvers support only l2 penalties. 'elasticnet' is only supported by the 'saga' solver. If 'none' (not supported by the liblinear solver), no regularization is applied.

   

   .. versionadded:: 0.19

   l1 penalty with SAGA solver (allowing 'multinomial' + L1)

   

   dual : bool, default=False

   Dual or primal formulation. Dual formulation is only implemented for l2 penalty with liblinear solver. Prefer dual=False when n_samples > n_features.

   

   tol : float, default=1e-4

   Tolerance for stopping criteria.

   

   C : float, default=1.0

   Inverse of regularization strength; must be a positive float.  Like in support vector machines, smaller values specify stronger regularization.

   

   fit_intercept : bool, default=True

   Specifies if a constant (a.k.a. bias or intercept) should be added to the decision function.

   

   intercept_scaling : float, default=1

   Useful only when the solver 'liblinear' is used  and self.fit_intercept is set to True. In this case, x becomes [x, self.intercept_scaling], i.e. a "synthetic" feature with constant value equal to intercept_scaling is appended to the instance vector.The intercept becomes ``intercept_scaling * synthetic_feature_weight``.

   

   Note! the synthetic feature weight is subject to l1/l2 regularization as all other features. To lessen the effect of regularization on synthetic feature weight (and therefore on the intercept) intercept_scaling has to be increased.

   

   class_weight : dict or 'balanced', default=None

   Weights associated with classes in the form ``{class_label: weight}``. If not given, all classes are supposed to have weight one.

   

   The "balanced" mode uses the values of y to automatically adjust  weights inversely proportional to class frequencies in the input data  as ``n_samples / (n_classes * np.bincount(y))``.

   

   Note that these weights will be multiplied with sample_weight (passed through the fit method) if sample_weight is specified.

   

   .. versionadded:: 0.17

   *class_weight='balanced'*

   

   random_state : int, RandomState instance, default=None Used when ``solver`` == 'sag', 'saga' or 'liblinear' to shuffle the data. See :term:`Glossary <random_state>` for details.

   

   solver : {'newton-cg', 'lbfgs', 'liblinear', 'sag', 'saga'}, \ default='lbfgs'

   

   Algorithm to use in the optimization problem.

   

   - For small datasets, 'liblinear' is a good choice, whereas 'sag' and 'saga' are faster for large ones.

   - For multiclass problems, only 'newton-cg', 'sag', 'saga' and 'lbfgs'  handle multinomial loss; 'liblinear' is limited to one-versus-rest  schemes.

   - 'newton-cg', 'lbfgs', 'sag' and 'saga' handle L2 or no penalty

   - 'liblinear' and 'saga' also handle L1 penalty

   - 'saga' also supports 'elasticnet' penalty

   - 'liblinear' does not support setting ``penalty='none'``

   

   Note that 'sag' and 'saga' fast convergence is only guaranteed on  features with approximately the same scale. You can preprocess the data with a scaler from sklearn.preprocessing. 参数

---------

处罚:{l1, l2,‘elasticnet’,‘没有’},默认=“l2”

用于指定在处罚中使用的规范。“newton-cg”,“sag”和“lbfgs”求解器只支持l2惩罚。“elasticnet”仅由“saga”求解器支持。如果“none”(liblinear求解器不支持),则不应用正则化。


. .versionadded:: 0.19

l1惩罚与SAGA求解器(允许“多项”+ l1)


bool,默认=False

双重或原始配方。对偶公式仅适用于l2罚用线性求解器。当n_samples > n_features时,preferred dual=False。


tol:浮动,默认=1e-4

停止标准的容忍度。


C: float, default=1.0

正则化强度的逆;必须是正浮点数。与支持向量机一样,值越小,正则化越强。


fit_intercept: bool,默认=True

指定一个常数(即偏差或拦截)是否应该添加到决策函数中。


intercept_scaling:浮动,默认=1

只有在使用“liblinear”求解器和self时才有用。fit_intercept设置为True。在这种情况下,x变成[x, self。intercept_scaling],即。一个常数值等于intercept_scaling的“合成”特性被附加到实例向量中。拦截变成' ' intercept_scaling * synthetic_feature_weight ' '。


注意!合成特征权重与所有其他特征一样,采用l1/l2正则化。为了减少正则化对合成特征权重的影响(因此对拦截的影响),必须增加intercept_scaling。


class_weight: dict或'balanced',默认为None

以' ' {class_label: weight} ' ' '形式关联类的权重。如果没有给出,所有类的权重都应该是1。


“平衡”模式使用y的值自动调整权重与输入数据中的类频率成反比,如' ' n_samples / (n_classes * np.bincount(y)) ' '。


注意,如果指定了sample_weight,那么这些权重将与sample_weight相乘(通过fit方法传递)。


. .versionadded:: 0.17

* class_weight = '平衡' *


random_state: int, RandomState instance, default=None,当' ' solver ' ' = 'sag', 'saga'或'liblinear'洗发数据时使用。详见:term: ' Glossary <random_state> '。</random_state>


解决:{‘newton-cg’,‘lbfgs’,‘liblinear’,“凹陷”,“传奇”},\默认=“lbfgs”


算法用于优化问题。


对于小数据集,“liblinear”是一个不错的选择,而“sag”和“saga”对于大数据集更快。

-对于多类问题,只有“newton-cg”、“sag”、“saga”和“lbfgs”处理多项损失;“liblinear”仅限于“一对二”方案。

- 'newton-cg', 'lbfgs', 'sag'和'saga'处理L2或没有处罚

-“liblinear”和“saga”也可以处理L1惩罚

-《英雄传奇》也支持《弹性网》的惩罚

- 'liblinear'不支持设置' ' penalty='none' ' '


请注意,“sag”和“saga”的快速收敛只能保证在大致相同规模的特性上。您可以使用sklearn.preprocessing中的scaler对数据进行预处理。

 .. versionadded:: 0.17

   Stochastic Average Gradient descent solver.

   .. versionadded:: 0.19

   SAGA solver.

   .. versionchanged:: 0.22

   The default solver changed from 'liblinear' to 'lbfgs' in 0.22.

   

   max_iter : int, default=100

   Maximum number of iterations taken for the solvers to converge.

   

   multi_class : {'auto', 'ovr', 'multinomial'}, default='auto'

   If the option chosen is 'ovr', then a binary problem is fit for each label. For 'multinomial' the loss minimised is the multinomial loss fit  across the entire probability distribution, *even when the data is binary*. 'multinomial' is unavailable when solver='liblinear'.  'auto' selects 'ovr' if the data is binary, or if solver='liblinear',  and otherwise selects 'multinomial'.

   

   .. versionadded:: 0.18

   Stochastic Average Gradient descent solver for 'multinomial' case.

   .. versionchanged:: 0.22

   Default changed from 'ovr' to 'auto' in 0.22.

   

   verbose : int, default=0

   For the liblinear and lbfgs solvers set verbose to any positive  number for verbosity.

   

   warm_start : bool, default=False

   When set to True, reuse the solution of the previous call to fit as initialization, otherwise, just erase the previous solution. Useless for liblinear solver. See :term:`the Glossary <warm_start>`.

   

   .. versionadded:: 0.17

   *warm_start* to support *lbfgs*, *newton-cg*, *sag*, *saga* solvers.

   

   n_jobs : int, default=None

   Number of CPU cores used when parallelizing over classes if  multi_class='ovr'". This parameter is ignored when the ``solver`` is set to 'liblinear' regardless of whether 'multi_class' is specified or not. ``None`` means 1 unless in a :obj:`joblib.parallel_backend`  context. ``-1`` means using all processors.

   See :term:`Glossary <n_jobs>` for more details.

   

   l1_ratio : float, default=None

   The Elastic-Net mixing parameter, with ``0 <= l1_ratio <= 1``. Only  used if ``penalty='elasticnet'``. Setting ``l1_ratio=0`` is equivalent to using ``penalty='l2'``, while setting ``l1_ratio=1`` is equivalent  to using ``penalty='l1'``. For ``0 < l1_ratio <1``, the penalty is a combination of L1 and L2.

. .versionadded:: 0.17

随机平均梯度下降求解器。

. .versionadded:: 0.19

SAGA solver。

. .versionchanged:: 0.22

在0.22中,默认求解器从“liblinear”更改为“lbfgs”。


max_iter: int,默认=100

使求解器收敛的最大迭代次数。


multi_class: {'auto', 'ovr', '多项'},默认='auto'

如果选择的选项是'ovr',那么每个标签都适合一个二进制问题。对于“多项”损失最小化是多项式损失适合整个概率分布,即使当数据是二进制*。当求解器='liblinear'时,不可用多项式。auto选择'ovr'如果数据是二进制的,或者solver='liblinear',否则选择'多项'。


. .versionadded:: 0.18

“多项式”情况的随机平均梯度下降求解器。

. .versionchanged:: 0.22

在0.22中默认从“ovr”改为“auto”。


int,默认=0

对于liblinear和lbfgs求解器,将冗长设置为任意正数。


warm_start: bool,默认=False

当设置为True时,重用前面调用的解决方案以适合初始化,否则就擦除前面的解决方案。对于线性求解器是没用的。参见:term: ' the Glossary <warm_start> '。</warm_start>


. .versionadded:: 0.17

*warm_start*支持*lbfgs*, *newton-cg*, *sag*, *saga*求解器。


n_jobs: int,默认=无

如果multi_class='ovr'",则在类上并行时使用的CPU核数。当' ' solver ' '被设置为'liblinear'时,不管'multi_class'是否被指定,这个参数都会被忽略。' ' None ' '表示1,除非在:obj: ' joblib.parallel_backend '上下文中。“-1”表示使用所有处理器。

有关更多细节,请参见:term: ' Glossary <n_jobs> '。</n_jobs>


l1_ratio: float, default=None

弹网混合参数``0 <= l1_ratio <= 1``。只在``penalty= ` elasticnet ``时使用。设置' ' l1_ratio=0 ' '等价于使用' ' penalty='l2' ' ',设置' ' l1_ratio=1 ' '等价于使用' ' penalty='l1' ' '。对于' ' 0 < l1_ratio <1 ' ',惩罚是L1和L2的组合。

   Attributes

   ----------

   

   classes_ : ndarray of shape (n_classes, )

   A list of class labels known to the classifier.

   

   coef_ : ndarray of shape (1, n_features) or (n_classes, n_features) Coefficient of the features in the decision function.

   

   `coef_` is of shape (1, n_features) when the given problem is binary.

   In particular, when `multi_class='multinomial'`, `coef_` corresponds to outcome 1 (True) and `-coef_` corresponds to outcome 0 (False).

   

   intercept_ : ndarray of shape (1,) or (n_classes,)

   Intercept (a.k.a. bias) added to the decision function.

   

   If `fit_intercept` is set to False, the intercept is set to zero.

   `intercept_` is of shape (1,) when the given problem is binary.  In particular, when `multi_class='multinomial'`, `intercept_`  corresponds to outcome 1 (True) and `-intercept_` corresponds to outcome 0 (False).

   

   n_iter_ : ndarray of shape (n_classes,) or (1, )

   Actual number of iterations for all classes. If binary or multinomial,  it returns only 1 element. For liblinear solver, only the maximum number of iteration across all classes is given.

   

   .. versionchanged:: 0.20

   

   In SciPy <= 1.0.0 the number of lbfgs iterations may exceed  ``max_iter``. ``n_iter_`` will now report at most ``max_iter``.

   

   See Also

   --------

   SGDClassifier : Incrementally trained logistic regression (when given the parameter ``loss="log"``).

   LogisticRegressionCV : Logistic regression with built-in cross validation.

   

   Notes

   -----

   The underlying C implementation uses a random number generator to select features when fitting the model. It is thus not uncommon,  to have slightly different results for the same input data. If  that happens, try with a smaller tol parameter.

   

   Predict output may not match that of standalone liblinear in certain cases. See :ref:`differences from liblinear <liblinear_differences>`  in the narrative documentation.

属性

----------


classes_:形状的ndarray

分类器已知的类标签列表。


coef_:决策函数中特征的形状(1,n_features)或(n_classes, n_features)系数的ndarray。


当给定的问题是二进制时,' coef_ '是形状(1,n_features)。

特别是,当“multi_class=”多项“”时,“coef_”对应结果1 (True),而“-coef_”对应结果0 (False)。


intercept_:形状(1,)或(n_classes,)的ndarray

在决策函数中加入截距(即偏差)。


如果' fit_intercept '设置为False,则拦截设置为零。

当给定的问题是二进制时,intercept_ '的形状是(1,)。特别是,当“multi_class=”多项“”时,“intercept_”对应结果1 (True),而“-intercept_”对应结果0 (False)。


n_iter_:形状(n_classes,)或(1,)的ndarray

所有类的实际迭代次数。如果是二项或多项,则只返回1个元素。对于线性求解器,只给出了所有类的最大迭代次数。


. .versionchanged:: 0.20


在SciPy <= 1.0.0中,lbfgs迭代次数可能超过' ' max_iter ' '。' ' n_iter_ ' '现在最多报告' ' max_iter ' '。


另请参阅

--------

增量训练逻辑回归(当给定参数' ' loss="log" ' ')。

逻辑回归cv:内置交叉验证的逻辑回归。



上一篇:centos6.3下haproxy+apache配置笔记


下一篇:LINUX常用命令(基础)