precision_recall_gain.prg.recall() - Code Metrics - crypdick/precision-recall-gain - Measure and Improve Code Quality continuously with Scrutinizer

precision_recall_gain.prg.recall() A
last analyzed 2024-02-17 01:58 UTC

↳ Parent: precision_recall_gain.prg

Complexity

Conditions

Size

Total Lines	3
Code Lines	3

Duplication

Lines	0
Ratio	0 %

Importance

Changes

Metric	Value
cc	2
eloc	3
nop	4
dl	0
loc	3
rs	10
c	0
b	0
f	0

from functools import partial

import numpy as np
from sklearn.metrics._base import _average_binary_score
from sklearn.metrics._ranking import _binary_clf_curve
from sklearn.utils.multiclass import type_of_target


def area_under_precision_recall_gain_score(
    y_true, y_score, *, average="macro", pos_label=1, sample_weight=None
):
    """Compute average precision (AP) from prediction scores.

    AP summarizes a precision-recall curve as the weighted mean of precisions
    achieved at each threshold, with the increase in recall from the previous
    threshold used as the weight:

    .. math::
        \\text{AP} = \\sum_n (R_n - R_{n-1}) P_n

    where :math:`P_n` and :math:`R_n` are the precision and recall at the nth
    threshold [1]_. This implementation is not interpolated and is different
    from computing the area under the precision-recall curve with the
    trapezoidal rule, which uses linear interpolation and can be too
    optimistic.

    Note: this implementation is restricted to the binary classification task
    or multilabel classification task.

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

    Parameters
    ----------
    y_true : ndarray of shape (n_samples,) or (n_samples, n_classes)
        True binary labels or binary label indicators.

    y_score : ndarray of shape (n_samples,) or (n_samples, n_classes)
        Target scores, can either be probability estimates of the positive
        class, confidence values, or non-thresholded measure of decisions
        (as returned by :term:`decision_function` on some classifiers).

    average : {'micro', 'samples', 'weighted', 'macro'} or None, \
            default='macro'
        If ``None``, the scores for each class are returned. Otherwise,
        this determines the type of averaging performed on the data:

        ``'micro'``:
            Calculate metrics globally by considering each element of the label
            indicator matrix as a label.
        ``'macro'``:
            Calculate metrics for each label, and find their unweighted
            mean.  This does not take label imbalance into account.
        ``'weighted'``:
            Calculate metrics for each label, and find their average, weighted
            by support (the number of true instances for each label).
        ``'samples'``:
            Calculate metrics for each instance, and find their average.

        Will be ignored when ``y_true`` is binary.

    pos_label : int or str, default=1
        The label of the positive class. Only applied to binary ``y_true``.
        For multilabel-indicator ``y_true``, ``pos_label`` is fixed to 1.

    sample_weight : array-like of shape (n_samples,), default=None
        Sample weights.

    Returns
    -------
    average_precision : float

    See Also
    --------
    roc_auc_score : Compute the area under the ROC curve.
    precision_recall_curve : Compute precision-recall pairs for different
        probability thresholds.

    Notes
    -----
    .. versionchanged:: 0.19
      Instead of linearly interpolating between operating points, precisions
      are weighted by the change in recall since the last operating point.

    References
    ----------
    .. [1] `Wikipedia entry for the Average precision
           <https://en.wikipedia.org/w/index.php?title=Information_retrieval&
           oldid=793358396#Average_precision>`_

    Examples
    --------
    >>> import numpy as np
    >>> from precision_recall_gain import average_precision_score
    >>> y_true = np.array([0, 0, 1, 1])
    >>> y_scores = np.array([0.1, 0.4, 0.35, 0.8])
    >>> average_precision_score(y_true, y_scores)
    0.83...
    """

    def _binary_uninterpolated_average_precision(
        y_true, y_score, pos_label=1, sample_weight=None
    ):
        precision_gain, recall_gain = precision_recall_gain_curve(
            y_true, y_score, pos_label=pos_label, sample_weight=sample_weight
        )
        # Return the step function integral
        # The following works because the last entry of precision is
        # guaranteed to be 1, as returned by precision_recall_curve
        # TODO compute integral correct?
        return -np.sum(np.diff(recall_gain) * np.array(precision_gain)[:-1])

    y_type = type_of_target(y_true)
    if y_type == "multilabel-indicator" and pos_label != 1:
        raise ValueError(
            "Parameter pos_label is fixed to 1 for "
            "multilabel-indicator y_true. Do not set "
            "pos_label or set pos_label to 1."
        )
    elif y_type == "binary":
        # Convert to Python primitive type to avoid NumPy type / Python str
        # comparison. See https://github.com/numpy/numpy/issues/6784
        present_labels = np.unique(y_true).tolist()
        if len(present_labels) == 2 and pos_label not in present_labels:
            raise ValueError(
                f"pos_label={pos_label} is not a valid label. It should be "
                f"one of {present_labels}"
            )
    average_precision = partial(
        _binary_uninterpolated_average_precision, pos_label=pos_label
    )
    # Average a binary metric for multilabel classification.
    average_precision = _average_binary_score(
        average_precision, y_true, y_score, average, sample_weight=sample_weight
    )
    return average_precision


def precision_recall_gain(precisions, recalls, proportion_of_positives):
    """
    Converts precision and recall into precision-gain and recall-gain.


    Parameters
    ----------
    proportion_of_positives: float. Proportion of positives. Termed π in the paper.
    precisions : ndarray
    recalls: ndarray
    """

    with np.errstate(divide="ignore", invalid="ignore"):
        prec_gain = (precisions - proportion_of_positives) / (
            (1 - proportion_of_positives) * precisions
        )
        rec_gain = (recalls - proportion_of_positives) / (
            (1 - proportion_of_positives) * recalls
        )

    return prec_gain, rec_gain


def precision_recall_gain_curve(y_true, probas_pred, pos_label=1, sample_weight=None):
    """Compute precision-recall pairs for different probability thresholds.

    Note: this implementation is restricted to the binary classification task.

    The precision is the ratio ``tp / (tp + fp)`` where ``tp`` is the number of
    true positives and ``fp`` the number of false positives. The precision is
    intuitively the ability of the classifier not to label as positive a sample
    that is negative.

    The recall is the ratio ``tp / (tp + fn)`` where ``tp`` is the number of
    true positives and ``fn`` the number of false negatives. The recall is
    intuitively the ability of the classifier to find all the positive samples.

    The last precision and recall values are 1. and 0. respectively and do not
    have a corresponding threshold. This ensures that the graph starts on the
    y axis.

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

    Parameters
    ----------
    y_true : ndarray of shape (n_samples,)
        True binary labels. If labels are not either {-1, 1} or {0, 1}, then
        pos_label should be explicitly given.

    probas_pred : ndarray of shape (n_samples,)
        Estimated probabilities or output of a decision function.

    pos_label : int or str, default=None
        The label of the positive class.
        When ``pos_label=None``, if y_true is in {-1, 1} or {0, 1},
        ``pos_label`` is set to 1, otherwise an error will be raised.

    sample_weight : array-like of shape (n_samples,), default=None
        Sample weights.

    Returns
    -------
    precision : ndarray of shape (n_thresholds + 1,)
        Precision values such that element i is the precision of
        predictions with score >= thresholds[i] and the last element is 1.

    recall : ndarray of shape (n_thresholds + 1,)
        Decreasing recall values such that element i is the recall of
        predictions with score >= thresholds[i] and the last element is 0.

    thresholds : ndarray of shape (n_thresholds,)
        Increasing thresholds on the decision function used to compute
        precision and recall. n_thresholds <= len(np.unique(probas_pred)).

    See Also
    --------
    plot_precision_recall_curve : Plot Precision Recall Curve for binary
        classifiers.
    PrecisionRecallDisplay : Precision Recall visualization.
    average_precision_score : Compute average precision from prediction scores.
    det_curve: Compute error rates for different probability thresholds.
    roc_curve : Compute Receiver operating characteristic (ROC) curve.

    Examples
    --------
    >>> import numpy as np
    >>> from precision_recall_gain import precision_recall_curve
    >>> y_true = np.array([0, 0, 1, 1])
    >>> y_scores = np.array([0.1, 0.4, 0.35, 0.8])
    >>> precision, recall, thresholds = precision_recall_curve(
    ...     y_true, y_scores)
    >>> precision
    array([0.66666667, 0.5       , 1.        , 1.        ])
    >>> recall
    array([1. , 0.5, 0.5, 0. ])
    >>> thresholds
    array([0.35, 0.4 , 0.8 ])

    """
    if pos_label != 1:
        raise NotImplementedError("Have not implemented non-binary targets")
    if sample_weight is not None:
        raise NotImplementedError

    # calc true and false poitives per binary classification thresh
    fps, tps, thresholds = _binary_clf_curve(
        y_true, probas_pred, pos_label=pos_label, sample_weight=sample_weight
    )

    precision = tps / (tps + fps)
    precision[np.isnan(precision)] = 0
    recall = tps / tps[-1]

    # stop when full recall attained
    # and reverse the outputs so recall is decreasing
    last_ind = tps.searchsorted(tps[-1])
    sl = slice(last_ind, None, -1)  # equivalent to slice [last_ind:None:-1]
    precision, recall, thresholds = (
        np.r_[precision[sl], 1],
        np.r_[recall[sl], 0],
        thresholds[sl],
    )

    # everything above is taken from sklearn.metrics._ranking.precision_recall_curve

    # logic taken from sklearn.metrics._ranking.det_curve
    # fns = tps[-1] - tps
    p_count = tps[-1]
    n_count = fps[-1]
    proportion_of_positives = p_count / n_count

    precision_gains, recall_gains = precision_recall_gain(
        precisions=precision,
        recalls=recall,
        proportion_of_positives=proportion_of_positives,
    )

    return precision_gains, recall_gains


"""
Source:
https://github.com/meeliskull/prg/blob/master/Python_package/prg/prg.py
"""


def precision(tp, fn, fp, tn):
    with np.errstate(divide="ignore", invalid="ignore"):
        return tp / (tp + fp)


def recall(tp, fn, fp, tn):
    with np.errstate(divide="ignore", invalid="ignore"):
        return tp / (tp + fn)


def precision_gain(tp, fn, fp, tn):

    """Calculates Precision Gain from the contingency table

    This function calculates Precision Gain from the entries of the contingency
    table: number of true positives (TP), false negatives (FN), false positives
    (FP), and true negatives (TN). More information on Precision-Recall-Gain
    curves and how to cite this work is available at
    http://www.cs.bris.ac.uk/~flach/PRGcurves/.
    """
    n_pos = tp + fn
    n_neg = fp + tn
    with np.errstate(divide="ignore", invalid="ignore"):
        prec_gain = 1.0 - (n_pos / n_neg) * (fp / tp)
    if np.alen(prec_gain) > 1:
        prec_gain[tn + fn == 0] = 0
    elif tn + fn == 0:
        prec_gain = 0
    return prec_gain


def recall_gain(tp, fn, fp, tn):

    """Calculates Recall Gain from the contingency table

    This function calculates Recall Gain from the entries of the contingency
    table: number of true positives (TP), false negatives (FN), false positives
    (FP), and true negatives (TN). More information on Precision-Recall-Gain
    curves and how to cite this work is available at
    http://www.cs.bris.ac.uk/~flach/PRGcurves/.

    Args:
        tp (float) or ([float]): True Positives
        fn (float) or ([float]): False Negatives
        fp (float) or ([float]): False Positives
        tn (float) or ([float]): True Negatives
    Returns:
        (float) or ([float])
    """
    n_pos = tp + fn
    n_neg = fp + tn
    with np.errstate(divide="ignore", invalid="ignore"):
        rg = 1.0 - (n_pos / n_neg) * (fn / tp)
    if np.alen(rg) > 1:
        rg[tn + fn == 0] = 1
    elif tn + fn == 0:
        rg = 1
    return rg


1		from functools import partial
2
3		import numpy as np
4		from sklearn.metrics._base import _average_binary_score
5		from sklearn.metrics._ranking import _binary_clf_curve
6		from sklearn.utils.multiclass import type_of_target
7
8
9		def area_under_precision_recall_gain_score(
10		y_true, y_score, *, average="macro", pos_label=1, sample_weight=None
11		):
12		"""Compute average precision (AP) from prediction scores.
13
14		AP summarizes a precision-recall curve as the weighted mean of precisions
15		achieved at each threshold, with the increase in recall from the previous
16		threshold used as the weight:
17
18		.. math::
19		\\text{AP} = \\sum_n (R_n - R_{n-1}) P_n
20
21		where :math:`P_n` and :math:`R_n` are the precision and recall at the nth
22		threshold [1]_. This implementation is not interpolated and is different
23		from computing the area under the precision-recall curve with the
24		trapezoidal rule, which uses linear interpolation and can be too
25		optimistic.
26
27		Note: this implementation is restricted to the binary classification task
28		or multilabel classification task.
29
30		Read more in the :ref:`User Guide <precision_recall_f_measure_metrics>`.
31
32		Parameters
33		----------
34		y_true : ndarray of shape (n_samples,) or (n_samples, n_classes)
35		True binary labels or binary label indicators.
36
37		y_score : ndarray of shape (n_samples,) or (n_samples, n_classes)
38		Target scores, can either be probability estimates of the positive
39		class, confidence values, or non-thresholded measure of decisions
40		(as returned by :term:`decision_function` on some classifiers).
41
42		average : {'micro', 'samples', 'weighted', 'macro'} or None, \
43		default='macro'
44		If ``None``, the scores for each class are returned. Otherwise,
45		this determines the type of averaging performed on the data:
46
47		``'micro'``:
48		Calculate metrics globally by considering each element of the label
49		indicator matrix as a label.
50		``'macro'``:
51		Calculate metrics for each label, and find their unweighted
52		mean. This does not take label imbalance into account.
53		``'weighted'``:
54		Calculate metrics for each label, and find their average, weighted
55		by support (the number of true instances for each label).
56		``'samples'``:
57		Calculate metrics for each instance, and find their average.
58
59		Will be ignored when ``y_true`` is binary.
60
61		pos_label : int or str, default=1
62		The label of the positive class. Only applied to binary ``y_true``.
63		For multilabel-indicator ``y_true``, ``pos_label`` is fixed to 1.
64
65		sample_weight : array-like of shape (n_samples,), default=None
66		Sample weights.
67
68		Returns
69		-------
70		average_precision : float
71
72		See Also
73		--------
74		roc_auc_score : Compute the area under the ROC curve.
75		precision_recall_curve : Compute precision-recall pairs for different
76		probability thresholds.
77
78		Notes
79		-----
80		.. versionchanged:: 0.19
81		Instead of linearly interpolating between operating points, precisions
82		are weighted by the change in recall since the last operating point.
83
84		References
85		----------
86		.. [1] `Wikipedia entry for the Average precision
87		<https://en.wikipedia.org/w/index.php?title=Information_retrieval&
88		oldid=793358396#Average_precision>`_
89
90		Examples
91		--------
92		>>> import numpy as np
93		>>> from precision_recall_gain import average_precision_score
94		>>> y_true = np.array([0, 0, 1, 1])
95		>>> y_scores = np.array([0.1, 0.4, 0.35, 0.8])
96		>>> average_precision_score(y_true, y_scores)
97		0.83...
98		"""
99
100		def _binary_uninterpolated_average_precision(
101		y_true, y_score, pos_label=1, sample_weight=None
102		):
103		precision_gain, recall_gain = precision_recall_gain_curve(
104		y_true, y_score, pos_label=pos_label, sample_weight=sample_weight
105		)
106		# Return the step function integral
107		# The following works because the last entry of precision is
108		# guaranteed to be 1, as returned by precision_recall_curve
109		# TODO compute integral correct?
110		return -np.sum(np.diff(recall_gain) * np.array(precision_gain)[:-1])
111
112		y_type = type_of_target(y_true)
113		if y_type == "multilabel-indicator" and pos_label != 1:
114		raise ValueError(
115		"Parameter pos_label is fixed to 1 for "
116		"multilabel-indicator y_true. Do not set "
117		"pos_label or set pos_label to 1."
118		)
119		elif y_type == "binary":
120		# Convert to Python primitive type to avoid NumPy type / Python str
121		# comparison. See https://github.com/numpy/numpy/issues/6784
122		present_labels = np.unique(y_true).tolist()
123		if len(present_labels) == 2 and pos_label not in present_labels:
124		raise ValueError(
125		f"pos_label={pos_label} is not a valid label. It should be "
126		f"one of {present_labels}"
127		)
128		average_precision = partial(
129		_binary_uninterpolated_average_precision, pos_label=pos_label
130		)
131		# Average a binary metric for multilabel classification.
132		average_precision = _average_binary_score(
133		average_precision, y_true, y_score, average, sample_weight=sample_weight
134		)
135		return average_precision
136
137
138		def precision_recall_gain(precisions, recalls, proportion_of_positives):
139		"""
140		Converts precision and recall into precision-gain and recall-gain.
141
142
143		Parameters
144		----------
145		proportion_of_positives: float. Proportion of positives. Termed π in the paper.
146		precisions : ndarray
147		recalls: ndarray
148		"""
149
150		with np.errstate(divide="ignore", invalid="ignore"):
151		prec_gain = (precisions - proportion_of_positives) / (
152		(1 - proportion_of_positives) * precisions
153		)
154		rec_gain = (recalls - proportion_of_positives) / (
155		(1 - proportion_of_positives) * recalls
156		)
157
158		return prec_gain, rec_gain
159
160
161		def precision_recall_gain_curve(y_true, probas_pred, pos_label=1, sample_weight=None):
162		"""Compute precision-recall pairs for different probability thresholds.
163
164		Note: this implementation is restricted to the binary classification task.
165
166		The precision is the ratio ``tp / (tp + fp)`` where ``tp`` is the number of
167		true positives and ``fp`` the number of false positives. The precision is
168		intuitively the ability of the classifier not to label as positive a sample
169		that is negative.
170
171		The recall is the ratio ``tp / (tp + fn)`` where ``tp`` is the number of
172		true positives and ``fn`` the number of false negatives. The recall is
173		intuitively the ability of the classifier to find all the positive samples.
174
175		The last precision and recall values are 1. and 0. respectively and do not
176		have a corresponding threshold. This ensures that the graph starts on the
177		y axis.
178
179		Read more in the :ref:`User Guide <precision_recall_f_measure_metrics>`.
180
181		Parameters
182		----------
183		y_true : ndarray of shape (n_samples,)
184		True binary labels. If labels are not either {-1, 1} or {0, 1}, then
185		pos_label should be explicitly given.
186
187		probas_pred : ndarray of shape (n_samples,)
188		Estimated probabilities or output of a decision function.
189
190		pos_label : int or str, default=None
191		The label of the positive class.
192		When ``pos_label=None``, if y_true is in {-1, 1} or {0, 1},
193		``pos_label`` is set to 1, otherwise an error will be raised.
194
195		sample_weight : array-like of shape (n_samples,), default=None
196		Sample weights.
197
198		Returns
199		-------
200		precision : ndarray of shape (n_thresholds + 1,)
201		Precision values such that element i is the precision of
202		predictions with score >= thresholds[i] and the last element is 1.
203
204		recall : ndarray of shape (n_thresholds + 1,)
205		Decreasing recall values such that element i is the recall of
206		predictions with score >= thresholds[i] and the last element is 0.
207
208		thresholds : ndarray of shape (n_thresholds,)
209		Increasing thresholds on the decision function used to compute
210		precision and recall. n_thresholds <= len(np.unique(probas_pred)).
211
212		See Also
213		--------
214		plot_precision_recall_curve : Plot Precision Recall Curve for binary
215		classifiers.
216		PrecisionRecallDisplay : Precision Recall visualization.
217		average_precision_score : Compute average precision from prediction scores.
218		det_curve: Compute error rates for different probability thresholds.
219		roc_curve : Compute Receiver operating characteristic (ROC) curve.
220
221		Examples
222		--------
223		>>> import numpy as np
224		>>> from precision_recall_gain import precision_recall_curve
225		>>> y_true = np.array([0, 0, 1, 1])
226		>>> y_scores = np.array([0.1, 0.4, 0.35, 0.8])
227		>>> precision, recall, thresholds = precision_recall_curve(
228		... y_true, y_scores)
229		>>> precision
230		array([0.66666667, 0.5 , 1. , 1. ])
231		>>> recall
232		array([1. , 0.5, 0.5, 0. ])
233		>>> thresholds
234		array([0.35, 0.4 , 0.8 ])
235
236		"""
237		if pos_label != 1:
238		raise NotImplementedError("Have not implemented non-binary targets")
239		if sample_weight is not None:
240		raise NotImplementedError
241
242		# calc true and false poitives per binary classification thresh
243		fps, tps, thresholds = _binary_clf_curve(
244		y_true, probas_pred, pos_label=pos_label, sample_weight=sample_weight
245		)
246
247		precision = tps / (tps + fps)
248		precision[np.isnan(precision)] = 0
249		recall = tps / tps[-1]
250
251		# stop when full recall attained
252		# and reverse the outputs so recall is decreasing
253		last_ind = tps.searchsorted(tps[-1])
254		sl = slice(last_ind, None, -1) # equivalent to slice [last_ind:None:-1]
255		precision, recall, thresholds = (
256		np.r_[precision[sl], 1],
257		np.r_[recall[sl], 0],
258		thresholds[sl],
259		)
260
261		# everything above is taken from sklearn.metrics._ranking.precision_recall_curve
262
263		# logic taken from sklearn.metrics._ranking.det_curve
264		# fns = tps[-1] - tps
265		p_count = tps[-1]
266		n_count = fps[-1]
267		proportion_of_positives = p_count / n_count
268
269		precision_gains, recall_gains = precision_recall_gain(
270		precisions=precision,
271		recalls=recall,
272		proportion_of_positives=proportion_of_positives,
273		)
274
275		return precision_gains, recall_gains
276
277
278		"""
279		Source:
280		https://github.com/meeliskull/prg/blob/master/Python_package/prg/prg.py
281		"""
282
283
284		def precision(tp, fn, fp, tn):
285		with np.errstate(divide="ignore", invalid="ignore"):
286		return tp / (tp + fp)
287
288
289		def recall(tp, fn, fp, tn):
290		with np.errstate(divide="ignore", invalid="ignore"):
291		return tp / (tp + fn)
292
293
294	View Code Duplication	def precision_gain(tp, fn, fp, tn):
		0 ignored issues – show Duplication introduced 2024-02-17 01:38 UTC by Report Bug Copy Issue Report This code seems to be duplicated in your project. Loading history...
295		"""Calculates Precision Gain from the contingency table
296
297		This function calculates Precision Gain from the entries of the contingency
298		table: number of true positives (TP), false negatives (FN), false positives
299		(FP), and true negatives (TN). More information on Precision-Recall-Gain
300		curves and how to cite this work is available at
301		http://www.cs.bris.ac.uk/~flach/PRGcurves/.
302		"""
303		n_pos = tp + fn
304		n_neg = fp + tn
305		with np.errstate(divide="ignore", invalid="ignore"):
306		prec_gain = 1.0 - (n_pos / n_neg) * (fp / tp)
307		if np.alen(prec_gain) > 1:
308		prec_gain[tn + fn == 0] = 0
309		elif tn + fn == 0:
310		prec_gain = 0
311		return prec_gain
312
313
314	View Code Duplication	def recall_gain(tp, fn, fp, tn):
		0 ignored issues – show Duplication introduced 2024-02-17 01:38 UTC by Report Bug Copy Issue Report This code seems to be duplicated in your project. Loading history...
315		"""Calculates Recall Gain from the contingency table
316
317		This function calculates Recall Gain from the entries of the contingency
318		table: number of true positives (TP), false negatives (FN), false positives
319		(FP), and true negatives (TN). More information on Precision-Recall-Gain
320		curves and how to cite this work is available at
321		http://www.cs.bris.ac.uk/~flach/PRGcurves/.
322
323		Args:
324		tp (float) or ([float]): True Positives
325		fn (float) or ([float]): False Negatives
326		fp (float) or ([float]): False Positives
327		tn (float) or ([float]): True Negatives
328		Returns:
329		(float) or ([float])
330		"""
331		n_pos = tp + fn
332		n_neg = fp + tn
333		with np.errstate(divide="ignore", invalid="ignore"):
334		rg = 1.0 - (n_pos / n_neg) * (fn / tp)
335		if np.alen(rg) > 1:
336		rg[tn + fn == 0] = 1
337		elif tn + fn == 0:
338		rg = 1
339		return rg
340

crypdick / precision-recall-gain

precision_recall_gain.prg.recall() A last analyzed 2024-02-17 01:58 UTC

Complexity

Size

Duplication

Importance

Duplication Side-by-Side

Filter issues like

precision_recall_gain.prg.recall() A
last analyzed 2024-02-17 01:58 UTC