fix: 모선 추론 점수 가중치 조정 — 100%는 DIRECT_PARENT_MATCH 전용

문제: china_bonus(15%) + prior(20%) 가산으로 일반 후보 23.6%가 100% 도달
- china_bonus: 0.15 → 0.05, 적용 조건: pre >= 0.30 → 0.50
- episode_prior: 0.10 → 0.05
- lineage_prior: 0.05 → 0.03
- label_prior: 0.10 → 0.07
- total_prior_cap: 0.20 → 0.10

결과: 일반 후보 최대 ~93% (라벨 있으면 ~98%), 100%는 직접 모선 일치만

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

prediction/algorithms/gear_parent_episode.py

@@ -23,10 +23,10 @@ _LABEL_PRIOR_WINDOW_DAYS = 30
 _CONTINUITY_SCORE_THRESHOLD = 0.45
 _MERGE_SCORE_THRESHOLD = 0.35
 _CENTER_DISTANCE_THRESHOLD_NM = 12.0
-_EPISODE_PRIOR_MAX = 0.10
-_LINEAGE_PRIOR_MAX = 0.05
-_LABEL_PRIOR_MAX = 0.10
-_TOTAL_PRIOR_CAP = 0.20
+_EPISODE_PRIOR_MAX = 0.05
+_LINEAGE_PRIOR_MAX = 0.03
+_LABEL_PRIOR_MAX = 0.07
+_TOTAL_PRIOR_CAP = 0.10
 
 
 def _clamp(value: float, floor: float = 0.0, ceil: float = 1.0) -> float:

prediction/algorithms/gear_parent_inference.py

@@ -51,8 +51,8 @@ _MIN_AUTO_PROMOTION_STABLE_CYCLES = 3
 _MIN_AUTO_PROMOTION_SCORE = 0.72
 _MIN_AUTO_PROMOTION_MARGIN = 0.15
 _MIN_REVIEW_REQUIRED_SCORE = 0.60
-_MIN_PREFIX_BONUS_SCORE = 0.30
-_CHINA_MMSI_PREFIX_BONUS = 0.15
+_MIN_PREFIX_BONUS_SCORE = 0.50
+_CHINA_MMSI_PREFIX_BONUS = 0.05
 _CHINA_MMSI_PREFIXES = ('412', '413')
 _TRACK_SUPPORT_POINT_TARGET = 12
 _TRACK_SUPPORT_SPAN_TARGET_MINUTES = 90.0