vertebrae.reports.recommendations

Simple recommendation rules for benchmark reports.

Functions

recommendation_for_extractor(score, stability, ...[, ...])	Compute a recommendation label for one extractor.
recommendations_for_benchmark(extractor_results[, ...])	Compute practitioner-facing recommendations for a benchmark.

Module Contents

vertebrae.reports.recommendations.recommendation_for_extractor(score, stability, weakest_class_score, target_type='single_label')

Compute a recommendation label for one extractor.

Parameters:

• score (float) – Primary overlap score.

• stability (Optional[Dict[str, Any]]) – Optional stability summary.

• weakest_class_score (Optional[float]) – Lowest per-class score when available.

• target_type (str)

Returns:

Recommendation label string.

Return type:

str

vertebrae.reports.recommendations.recommendations_for_benchmark(extractor_results, quality_tolerance=None)

Compute practitioner-facing recommendations for a benchmark.

Parameters:

• extractor_results (List[Any]) – Evaluated extractor results.

• quality_tolerance (Optional[float])

Returns:

Recommendation messages for the report executive summary.

Return type:

List[str]