src.dackar.RCA.pm_compliance.effectiveness_analyzer
===================================================

.. py:module:: src.dackar.RCA.pm_compliance.effectiveness_analyzer

.. autoapi-nested-parse::

   PMEffectivenessAnalyzer — as-found trend signals (advisory, data-quality limited).



Functions
---------

.. autoapisummary::

   src.dackar.RCA.pm_compliance.effectiveness_analyzer.analyze_degradation
   src.dackar.RCA.pm_compliance.effectiveness_analyzer.compute_pm_found_defect_rate
   src.dackar.RCA.pm_compliance.effectiveness_analyzer.collect_as_found_from_checks
   src.dackar.RCA.pm_compliance.effectiveness_analyzer.collect_as_found_from_rows


Module Contents
---------------

.. py:function:: analyze_degradation(as_found_texts, data_dir = None)

   Return *degradation_trend* label based on as-found text.

   When *data_dir* points to the DACKAR data directory, uses the curated
   ``health_status_keywords_negative/positive.csv`` vocabulary.  Falls back to
   hardcoded stems when *data_dir* is ``None`` or the files are absent.

   Classification logic (in priority order):
   1. No non-empty texts → ``"unknown"``
   2. Improving terms match AND no degrading terms match → ``"improving"``
   3. Degrading terms match → ``"degrading"``
   4. Otherwise → ``"stable"``


.. py:function:: compute_pm_found_defect_rate(rows, data_dir = None)

   Return fraction of PM executions with as-found data that recorded a defect.

   Iterates *all* export rows (not cycle-limited) to give a statistical signal
   over the full lookback history.  Returns ``None`` when no rows carry as-found
   data (avoids a misleading 0.0 rate).

   A row is counted as a defect when its as-found text classifies as
   ``"degrading"`` per ``analyze_degradation``.


.. py:function:: collect_as_found_from_checks(checks)

   Use ``details`` (as-found) from checks when available.


.. py:function:: collect_as_found_from_rows(rows, max_cycles = None)

   Prefer structured as-found columns from export rows.

   When *max_cycles* is set, only the most-recent N rows (by ``completed_date``)
   are considered — honouring ``PMComplianceConfig.effectiveness_lookback_cycles``.