ML之sklearn:sklearn.metrics中常用的函数参数(比如confusion_matrix等 )解释及其用法说明之详细攻略

ML之sklearn:sklearn.metrics中常用的函数参数(比如confusion_matrix等 )解释及其用法说明之详细攻略

 

 

 

目录

sklearn.metrics中常用的函数参数

confusion_matrix


 

 

推荐文章
ML:分类预测问题中评价指标(ER/混淆矩阵P-R-F1/ROC-AUC/RP/mAP)简介、使用方法、代码实现、案例应用之详细攻略
CNN之性能指标:卷积神经网络中常用的性能指标(IOU/AP/mAP、混淆矩阵)简介、使用方法之详细攻略

sklearn.metrics中常用的函数参数

confusion_matrix函数解释

返回值:混淆矩阵,其第i行和第j列条目表示真实标签为第i类、预测标签为第j类的样本数。

                                           预测
                     0                    1
真实    0  
    1  

 

def confusion_matrix Found at: sklearn.metrics._classification

@_deprecate_positional_args
def confusion_matrix(y_true, y_pred, *, labels=None, sample_weight=None,  normalize=None):
    """Compute confusion matrix to evaluate the accuracy of a classification.
    
    By definition a confusion matrix :math:`C` is such that :math:`C_{i, j}` is equal to the number of observations known to be in group :math:`i` and predicted to be in group :math:`j`.
    
    Thus in binary classification, the count of true negatives is
    :math:`C_{0,0}`, false negatives is :math:`C_{1,0}`, true positives is
    :math:`C_{1,1}` and false positives is :math:`C_{0,1}`.
    
    Read more in the :ref:`User Guide <confusion_matrix>`.
    
    Parameters
    ----------
    y_true : array-like of shape (n_samples,) Ground truth (correct) target values.
    y_pred : array-like of shape (n_samples,) Estimated targets as returned by a classifier.
    labels : array-like of shape (n_classes), default=None.  List of labels to index the matrix. This may be used to reorder
    or select a subset of labels.  If ``None`` is given, those that appear at least once in ``y_true`` or ``y_pred`` are used in sorted order.
    
    sample_weight : array-like of shape (n_samples,), default=None. Sample weights.
    
    .. versionadded:: 0.18
    
    normalize : {'true', 'pred', 'all'}, default=None. Normalizes confusion matrix over the true (rows), predicted (columns)
    conditions or all the population. If None, confusion matrix will not be normalized.
    
    Returns
    -------
    C : ndarray of shape (n_classes, n_classes)
    Confusion matrix whose i-th row and j-th column entry indicates the number of samples with true label being i-th class and prediced label being j-th class.
    
    References
    ----------
    .. [1] `Wikipedia entry for the Confusion matrix <https://en.wikipedia.org/wiki/Confusion_matrix>`_  (Wikipedia and other references may use a different convention for axes)

在:sklear. metrics._classification找到的def confusion_matrix

@_deprecate_positional_args
defconfusion_matrix (y_true, y_pred, *, label =None, sample_weight=None, normalize= None):
计算混淆矩阵来评估分类的准确性

根据定义,一个混淆矩阵:math: ' C '是这样的:math: ' C_{i, j} '等于已知在:math: ' i '组和预测在:math: ' j '组的观测数。

因此,在二元分类法中,true negatives的数量是
    :math:`C_{0,0}`, false negatives is :math:`C_{1,0}`, true positives is
    :math:`C_{1,1}` and false positives is :math:`C_{0,1}`.

更多信息见:ref: ' User Guide <confusion_matrix> '。</confusion_matrix>

参数
----------
y_true:类数组形状(n_samples,) Ground truth (correct)目标值。
y_pred:分类器返回的估计目标的类数组形状(n_samples,)。
标签:类数组形状(n_classes),默认=无。索引矩阵的标签列表。这可以用于重新排序
或者选择标签的子集。如果给出了' ' None ' ',则在' ' y_true ' '或' ' y_pred ' '中至少出现一次的值将按排序顺序使用。

sample_weight:类似数组的形状(n_samples,),默认=None。样本权重。

. .versionadded:: 0.18

{'true', 'pred', 'all'}, default=None。对真实(行)、预测(列)的混淆矩阵进行规范化
条件或所有的人口。如果没有,混淆矩阵将不会被标准化。

返回
-------
C:形状的ndarray (n_classes, n_classes)
第i行和第j列项表示真标签样本个数为第i类,谓词标签样本个数为第j类的混淆矩阵。
&nbsp; &nbsp;&nbsp;
引用
----------
. .[1] '用于混淆矩阵的维基百科条目<https: en.wikipedia.org="" wiki="" confusion_matrix=""> ' _(维基百科和其他引用可能对轴使用不同的约定)</https:>

  Examples
    --------
    >>> from sklearn.metrics import confusion_matrix
    >>> y_true = [2, 0, 2, 2, 0, 1]
    >>> y_pred = [0, 0, 2, 2, 0, 2]
    >>> confusion_matrix(y_true, y_pred)
    array([[2, 0, 0],
    [0, 0, 1],
    [1, 0, 2]])
    
    >>> y_true = ["cat", "ant", "cat", "cat", "ant", "bird"]
    >>> y_pred = ["ant", "ant", "cat", "cat", "ant", "cat"]
    >>> confusion_matrix(y_true, y_pred, labels=["ant", "bird", "cat"])
    array([[2, 0, 0],
    [0, 0, 1],
    [1, 0, 2]])
    
    In the binary case, we can extract true positives, etc as follows:
    
    >>> tn, fp, fn, tp = confusion_matrix([0, 1, 0, 1], [1, 1, 1, 0]).ravel()
    >>> (tn, fp, fn, tp)
    (0, 2, 1, 1)
 
    """
    y_type, y_true, y_pred = _check_targets(y_true, y_pred)
    if y_type not in ("binary", "multiclass"):
        raise ValueError("%s is not supported" % y_type)
    if labels is None:
        labels = unique_labels(y_true, y_pred)
    else:
        labels = np.asarray(labels)
        n_labels = labels.size
        if n_labels == 0:
            raise ValueError("'labels' should contains at least one label.")
        elif y_true.size == 0:
            return np.zeros((n_labels, n_labels), dtype=np.int)
        elif np.all([l not in y_true for l in labels]):
            raise ValueError("At least one label specified must be in y_true")
    if sample_weight is None:
        sample_weight = np.ones(y_true.shape[0], dtype=np.int64)
    else:
        sample_weight = np.asarray(sample_weight)
    check_consistent_length(y_true, y_pred, sample_weight)
    if normalize not in ['true', 'pred', 'all', None]:
        raise ValueError("normalize must be one of {'true', 'pred', "
            "'all', None}")
    n_labels = labels.size
    label_to_ind = {y:x for x, y in enumerate(labels)}
    # convert yt, yp into index
    y_pred = np.array([label_to_ind.get(x, n_labels + 1) for x in y_pred])
    y_true = np.array([label_to_ind.get(x, n_labels + 1) for x in y_true])
    # intersect y_pred, y_true with labels, eliminate items not in labels
    ind = np.logical_and(y_pred < n_labels, y_true < n_labels)
    y_pred = y_pred[ind]
    y_true = y_true[ind] # also eliminate weights of eliminated items
    sample_weight = sample_weight[ind]
    # Choose the accumulator dtype to always have high precision
    if sample_weight.dtype.kind in {'i', 'u', 'b'}:
        dtype = np.int64
    else:
        dtype = np.float64
    cm = coo_matrix((sample_weight, (y_true, y_pred)), shape=(n_labels, 
     n_labels), dtype=dtype).toarray()
    with np.errstate(all='ignore'):
        if normalize == 'true':
            cm = cm / cm.sum(axis=1, keepdims=True)
        elif normalize == 'pred':
            cm = cm / cm.sum(axis=0, keepdims=True)
        elif normalize == 'all':
            cm = cm / cm.sum()
        cm = np.nan_to_num(cm)
    return cm
 

 

 

 

 

 

 

 

 

 

 

 

 

©️2020 CSDN 皮肤主题: 代码科技 设计师:Amelia_0503 返回首页
实付 29.90元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、C币套餐、付费专栏及课程。

余额充值