feat(sp6): 4-field weighted scoring module

2026-06-20 18:30:05 -06:00
parent f4910fd94d
commit 3432c76561
2 changed files with 221 additions and 0 deletions
@@ -0,0 +1,96 @@
+"""Transparent 4-field weighted scoring for Inbox candidate pairs.
+
+SP6 T5.
+
+Weights (sum to 100):
+  - patient control number match (exact, normalized)        40
+  - service date proximity (±3 days linear decay)           25
+  - charge amount proximity (±10% linear decay)             20
+  - provider NPI exact match                                15
+"""
+from __future__ import annotations
+
+from dataclasses import dataclass
+from datetime import date
+from decimal import Decimal
+from typing import Protocol
+
+
+class _ScorableClaim(Protocol):
+    patient_control_number: str
+    service_date_from: date | None
+    charge_amount: Decimal
+    provider_npi: str | None
+
+
+class _ScorableRemit(Protocol):
+    patient_control_number: str
+    service_date: date | None
+    charge_amount: Decimal
+    rendering_provider_npi: str | None
+
+
+@dataclass(frozen=True)
+class ScoreBreakdown:
+    patient: float
+    date: float
+    amount: float
+    provider: float
+
+    @property
+    def total(self) -> int:
+        return round(
+            self.patient * 40
+            + self.date * 25
+            + self.amount * 20
+            + self.provider * 15
+        )
+
+    @property
+    def tier(self) -> str:
+        if self.total >= 75:
+            return "strong"
+        if self.total >= 50:
+            return "weak"
+        return "hidden"
+
+
+def _norm_pcn(s: str) -> str:
+    return (s or "").strip().lower().lstrip("0")
+
+
+def _date_score(claim_date: date | None, remit_date: date | None) -> float:
+    if claim_date is None or remit_date is None:
+        return 0.0
+    delta = abs((claim_date - remit_date).days)
+    if delta > 3:
+        return 0.0
+    return 1.0 - (delta / 3.0)
+
+
+def _amount_score(claim_amt: Decimal, remit_amt: Decimal) -> float:
+    if claim_amt <= 0:
+        return 0.0
+    diff = abs(claim_amt - remit_amt)
+    threshold = claim_amt * Decimal("0.10")
+    if diff > threshold:
+        return 0.0
+    # Normalize diff within threshold to 0..1
+    return float(Decimal("1.0") - (diff / threshold))
+
+
+def _provider_score(claim_npi: str | None, remit_npi: str | None) -> float:
+    if not claim_npi or not remit_npi:
+        return 0.0
+    return 1.0 if claim_npi.strip() == remit_npi.strip() else 0.0
+
+
+def score_pair(claim: _ScorableClaim, remit: _ScorableRemit) -> ScoreBreakdown:
+    claim_pcn = _norm_pcn(claim.patient_control_number)
+    remit_pcn = _norm_pcn(remit.patient_control_number)
+    return ScoreBreakdown(
+        patient=1.0 if claim_pcn and claim_pcn == remit_pcn else 0.0,
+        date=_date_score(claim.service_date_from, remit.service_date),
+        amount=_amount_score(claim.charge_amount, remit.charge_amount),
+        provider=_provider_score(claim.provider_npi, remit.rendering_provider_npi),
+    )
@@ -0,0 +1,125 @@
+"""TDD: 4-field weighted scoring for Inbox candidate pairs.
+
+Plan: docs/superpowers/plans/2026-06-20-cyclone-workflow-automation.md
+Task 5 (T5).
+"""
+from __future__ import annotations
+
+from dataclasses import dataclass
+from datetime import date
+from decimal import Decimal
+
+import pytest
+
+from cyclone.scoring import score_pair
+
+
+@dataclass
+class FakeClaim:
+    patient_control_number: str = "PCN-001"
+    service_date_from: date | None = date(2026, 6, 10)
+    charge_amount: Decimal = Decimal("100.00")
+    provider_npi: str | None = "1234567890"
+
+
+@dataclass
+class FakeRemit:
+    patient_control_number: str = "PCN-001"
+    service_date: date | None = date(2026, 6, 10)
+    charge_amount: Decimal = Decimal("100.00")
+    rendering_provider_npi: str | None = "1234567890"
+
+
+def test_perfect_match_scores_100():
+    s = score_pair(FakeClaim(), FakeRemit())
+    assert s.total == 100
+    assert s.patient == 1.0
+    assert s.date == 1.0
+    assert s.amount == 1.0
+    assert s.provider == 1.0
+
+
+def test_date_within_3_days_scores_partial():
+    s = score_pair(
+        FakeClaim(service_date_from=date(2026, 6, 10)),
+        FakeRemit(service_date=date(2026, 6, 12)),
+    )
+    assert s.date == pytest.approx(1 - 2/3)   # 0.333
+
+
+def test_date_beyond_3_days_scores_zero():
+    s = score_pair(
+        FakeClaim(service_date_from=date(2026, 6, 10)),
+        FakeRemit(service_date=date(2026, 6, 14)),
+    )
+    assert s.date == 0.0
+
+
+def test_amount_within_10pct_scores_partial():
+    s = score_pair(
+        FakeClaim(charge_amount=Decimal("100.00")),
+        FakeRemit(charge_amount=Decimal("108.00")),
+    )
+    assert s.amount == pytest.approx(1 - 8/10)
+
+
+def test_amount_beyond_10pct_scores_zero():
+    s = score_pair(
+        FakeClaim(charge_amount=Decimal("100.00")),
+        FakeRemit(charge_amount=Decimal("120.00")),
+    )
+    assert s.amount == 0.0
+
+
+def test_patient_normalized_exact_match():
+    s = score_pair(
+        FakeClaim(patient_control_number="PCN-001"),
+        FakeRemit(patient_control_number=" pcn-001 "),
+    )
+    assert s.patient == 1.0
+
+
+def test_patient_mismatch_scores_zero():
+    s = score_pair(
+        FakeClaim(patient_control_number="PCN-001"),
+        FakeRemit(patient_control_number="PCN-002"),
+    )
+    assert s.patient == 0.0
+
+
+def test_missing_provider_npi_scores_zero():
+    s = score_pair(
+        FakeClaim(provider_npi=None),
+        FakeRemit(rendering_provider_npi="1234567890"),
+    )
+    assert s.provider == 0.0
+
+
+def test_missing_service_date_scores_zero():
+    s = score_pair(
+        FakeClaim(service_date_from=None),
+        FakeRemit(service_date=date(2026, 6, 10)),
+    )
+    assert s.date == 0.0
+
+
+def test_threshold_boundaries():
+    """Tier boundaries: ≥75 strong, 50–74 weak, <50 hidden."""
+    # All fields 0 except amount = 0.75 (exact) → 15 pts → total 15 → hidden
+    s = score_pair(
+        FakeClaim(
+            patient_control_number="X",        # 0
+            service_date_from=date(2026, 6, 10),
+            charge_amount=Decimal("100.00"),
+            provider_npi="1234567890",
+        ),
+        FakeRemit(
+            patient_control_number="Y",        # 0
+            service_date=date(2026, 6, 10),
+            charge_amount=Decimal("100.00"),
+            rendering_provider_npi="1234567890",
+        ),
+    )
+    # patient=0, date=1, amount=1, provider=1 → 0+25+20+15 = 60 → weak tier
+    assert 50 <= s.total < 75   # weak tier
+    assert s.tier == "weak"