F_b score

[rsuwaileh]

Hey,

Is it possible to add F_beta to the list of measures? This is in case I want to weight P or R more than the other?
F_b = (1 + b^2) * (precision * recall) / ((b^2 * precision) + recall)

seqeval/seqeval/metrics/sequence_labeling.py

Line 124 in 445d997

F1 = 2 * (precision * recall) / (precision + recall)

score = (1 + b**2) * p * r / ((b**2 * p) + r) if p + r > 0 else 0

seqeval/seqeval/metrics/sequence_labeling.py

Line 149 in 445d997

score = 2 * p * r / (p + r) if p + r > 0 else 0

[Hironsan]

As of v1.0.0, precision_recall_fscore_support function supports beta. You can calculate F-beta easily.

seqeval/seqeval/metrics/v1.py

Lines 100 to 109 in 91215f5

	def precision_recall_fscore_support(y_true: List[List[str]],
	y_pred: List[List[str]],
	*,
	average: Optional[str] = None,
	warn_for=('precision', 'recall', 'f-score'),
	beta: float = 1.0,
	sample_weight=None,
	zero_division: str = 'warn',
	scheme: Type[Token] = None,
	suffix: bool = False) -> SCORES: