Metrics evaluator

fyt.core.metrics_evaluator

Metrics evaluation for trained models.

Extracted from Trainer to follow the Single Responsibility Principle.

MetricsEvaluator

Computes evaluation metrics for a trained model.

Parameters:

Name	Type	Description	Default
metrics	list[MetricType | str]	List of metrics to compute.	required
metric_averaging	str	Averaging strategy for precision/recall/f1. "auto" resolves to "binary" for two-class data and "weighted" otherwise.	'auto'

compute(model, data_split, random_state)

Compute evaluation metrics on the test set.

Parameters:

Name	Type	Description	Default
model	Any	Trained model with predict() and optionally predict_proba().	required
data_split	DataSplit	Data split containing test data.	required
random_state	int	Random state seed used for this evaluation.	required

Returns:

Type	Description
MetricResults	MetricResults containing all computed metrics.