Metrics#

Model-agnostic probabilistic evaluation metrics.

Functions

`log_loss`	Mean binary log loss; lower is better.
`brier_score`	Binary Brier score; lower is better.
`expected_calibration_error`	Equal-width expected calibration error; lower is better.
`bootstrap_ci`	Percentile bootstrap confidence interval for a statistic.
`classification_report`	Return a JSON-friendly probabilistic evaluation summary.

log_loss(y_true, y_prob)[source]#: Mean binary log loss; lower is better.

brier_score(y_true, y_prob)[source]#: Binary Brier score; lower is better.

expected_calibration_error(y_true, y_prob, bins=10)[source]#: Equal-width expected calibration error; lower is better.

bootstrap_ci(values, statistic=<function mean>, confidence=0.95, n_boot=2000, rng=None)[source]#: Percentile bootstrap confidence interval for a statistic.

classification_report(y_true, y_prob, bins=10)[source]#: Return a JSON-friendly probabilistic evaluation summary.