def test_precision(caplog):
"""Unit test of precision_scorer"""
caplog.set_level(logging.INFO)
golds = np.array([0, 1, 0, 1, 0, 1])
preds = np.array([0, 0, 0, 0, 0, 1])
metric_dict = precision_scorer(golds, None, preds, pos_label=1)
assert isequal(metric_dict, {"precision": 1})
metric_dict = precision_scorer(golds, None, preds, pos_label=0)
assert isequal(metric_dict, {"precision": 0.6})
def fbeta_scorer(golds, probs, preds, uids=None, pos_label=1, beta=1):
"""F-beta score is the weighted harmonic mean of precision and recall.
:param golds: Ground truth (correct) target values.
:type golds: 1-d np.array
:param probs: Predicted target probabilities. (Not used!)
:type probs: k-d np.array
:param preds: Predicted target values.
:type preds: 1-d np.array
:param uids: Unique ids.
:type uids: list
:param pos_label: The positive class label, defaults to 1
:param pos_label: int, optional
:param beta: Weight of precision in harmonic mean, defaults to 1
:param beta: float, optional
:return: F-beta score.
:rtype: dict
"""
precision = precision_scorer(golds, probs, preds, uids,
pos_label)["precision"]
recall = recall_scorer(golds, probs, preds, uids, pos_label)["recall"]
fbeta = ((1 + beta**2) * (precision * recall) /
((beta**2 * precision) + recall) if
(beta**2 * precision) + recall > 0 else 0.0)
return {f"f{beta}": fbeta}
def fbeta_scorer(
golds: ndarray,
probs: Optional[ndarray],
preds: ndarray,
uids: Optional[List[str]] = None,
pos_label: int = 1,
beta: int = 1,
) -> Dict[str, float]:
"""F-beta score is the weighted harmonic mean of precision and recall.
Args:
golds(ndarray): Ground truth values.
probs(ndarray or None): Predicted probabilities.
preds(ndarray): Predicted values.
uids(list, optional): Unique ids, defaults to None.
pos_label(int, optional): The positive class label, defaults to 1.
beta(float, optional): Weight of precision in harmonic mean, defaults to 1.
Returns:
dict: F-beta score.
"""
precision = precision_scorer(golds, probs, preds, uids,
pos_label)["precision"]
recall = recall_scorer(golds, probs, preds, uids, pos_label)["recall"]
fbeta = ((1 + beta**2) * (precision * recall) /
((beta**2 * precision) + recall) if
(beta**2 * precision) + recall > 0 else 0.0)
return {f"f{beta}": fbeta}
def fbeta_scorer(
golds: ndarray,
probs: Optional[ndarray],
preds: ndarray,
uids: Optional[List[str]] = None,
pos_label: int = 1,
beta: int = 1,
) -> Dict[str, float]:
"""F-beta score is the weighted harmonic mean of precision and recall.
Args:
golds: Ground truth values.
probs: Predicted probabilities.
preds: Predicted values.
uids: Unique ids, defaults to None.
pos_label: The positive class label, defaults to 1.
beta: Weight of precision in harmonic mean, defaults to 1.
Returns:
F-beta score.
"""
# Convert probabilistic label to hard label
if len(golds.shape) == 2:
golds = prob_to_pred(golds)
precision = precision_scorer(golds, probs, preds, uids,
pos_label)["precision"]
recall = recall_scorer(golds, probs, preds, uids, pos_label)["recall"]
fbeta = ((1 + beta**2) * (precision * recall) /
((beta**2 * precision) + recall) if
(beta**2 * precision) + recall > 0 else 0.0)
return {f"f{beta}": fbeta}
def test_precision(caplog):
"""Unit test of precision_scorer"""
caplog.set_level(logging.INFO)
golds = np.array([0, 1, 0, 1, 0, 1])
gold_probs = np.array([[0.6, 0.4], [0.1, 0.9], [0.7, 0.3], [0.2, 0.8],
[0.9, 0.1], [0.4, 0.6]])
preds = np.array([0, 0, 0, 0, 0, 1])
metric_dict = precision_scorer(golds, None, preds, pos_label=1)
assert isequal(metric_dict, {"precision": 1})
metric_dict = precision_scorer(golds, None, preds, pos_label=0)
assert isequal(metric_dict, {"precision": 0.6})
metric_dict = precision_scorer(gold_probs, None, preds, pos_label=1)
assert isequal(metric_dict, {"precision": 1})
metric_dict = precision_scorer(gold_probs, None, preds, pos_label=0)
assert isequal(metric_dict, {"precision": 0.6})
def test_precision(caplog):
"""Unit test of precision_scorer."""
caplog.set_level(logging.INFO)
metric_dict = precision_scorer(GOLDS, PROBS, PREDS, pos_label=1)
assert isequal(metric_dict, {"precision": 1})
metric_dict = precision_scorer(GOLDS, None, PREDS, pos_label=1)
assert isequal(metric_dict, {"precision": 1})
metric_dict = precision_scorer(GOLDS, None, PREDS, pos_label=0)
assert isequal(metric_dict, {"precision": 0.6})
metric_dict = precision_scorer(PROB_GOLDS, PROBS, PREDS, pos_label=1)
assert isequal(metric_dict, {"precision": 1})
metric_dict = precision_scorer(PROB_GOLDS, None, PREDS, pos_label=1)
assert isequal(metric_dict, {"precision": 1})
metric_dict = precision_scorer(PROB_GOLDS, None, PREDS, pos_label=0)
assert isequal(metric_dict, {"precision": 0.6})