junte/Model
1
0
Fork 0
mirror of https://github.com/Hestia-Homes/Model.git synced 2026-07-27 23:35:01 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Tighten pashub accuracy ratchets to 0.51 / 0.625 🟪

Browse source 
#1602 (floor exposure flags) + #1603 (alternative walls + extension floor
construction) move the Guinness 205 cohort within-0.5 50.2% -> 51.2% and MAE
0.653 -> 0.624. Lock the gain: floor 0.50 -> 0.51, ceiling 0.653 -> 0.625.
#1601 note updated from HELD to RESOLVED (keep-raw).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
Khalim Conn-Kowlessar 2026-07-15 15:46:06 +00:00
parent 5324ae0e57
commit 6fd0393a4e
1 changed files with 18 additions and 11 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

29
backend/documents_parser/tests/test_pashub_sap_accuracy.py
View file

@ -92,12 +92,17 @@ _MANIFEST_PATH = _FIXTURES_DIR / "manifest.json"
# + bath counts never threaded into `sap_heating`; #1599 floor label "Ground
# Floor" normalized to the calculator's "Ground floor" so the §5(12) suspended-
# timber infiltration term fires; #1600 water-heating "None" coded to the RdSAP
# §10.7 WHC 999 default instead of inheriting the gas combi). Observed 50.2%
# (103/205) — the campaign's largest move. All three are silent-drop fixes the
# gov-API/Elmhurst paths already apply; verified against Khalim's ground truth
# (16 Stillwater 78.2 -> 75.7 vs 76; 58 Hackle 67.7 -> ~58 vs 59). #1601 (upper-
# floor +0.25 m) is HELD — it empirically over-corrects this cohort.
_MIN_WITHIN_HALF: float = 0.50
# §10.7 WHC 999 default instead of inheriting the gas combi). All three are
# silent-drop fixes the gov-API/Elmhurst paths already apply; verified against
# Khalim's ground truth (16 Stillwater 78.2 -> 75.7 vs 76; 58 Hackle 67.7 ->
# ~58 vs 59). Ratcheted 0.50 -> 0.51 by #1602 (semi-exposed / exposed floor
# exposure flags -> RdSAP §5.14 U=0.7 / §5.13 Table 20) + #1603 (alternative-
# wall sub-areas + extension floor construction now threaded): observed 51.2%
# (105/205). #1601 (upper-floor +0.25 m) RESOLVED as keep-raw — adjudicated
# against verified truth it over-corrects every verified dwelling (raw beats
# +0.25 on all 7); pashub surveys internal dims and its upper-ground height
# gap is modal 0.00, so the joist void must not be added.
_MIN_WITHIN_HALF: float = 0.51
# Ratcheted 2.71 -> 2.64 by #1590 bug 2 (roof "Insulation At: None" -> explicit
# zero thickness): observed MAE 2.632 — the two ~7-SAP roof over-raters
# (507665533138 / 507670893756) correcting is exactly 14/205 ≈ 0.07 of MAE.
@ -158,11 +163,13 @@ _MIN_WITHIN_HALF: float = 0.50
# shower + bath counts; #1599 floor label "Ground Floor" -> "Ground floor" so the
# §5(12) suspended-timber infiltration fires; #1600 water-heating "None" -> RdSAP
# §10.7 WHC 999 — see the within-0.5 note above), which SUPERSEDES the #1592
# loosening: observed MAE 0.6517 on the merged tree (with the water-only DHW fix,
# 16 Bingley still +6.818). The three silent-drop fixes the gov-API/Elmhurst
# paths already apply; the electric-shower drop (100/205) was the dominant lever.
# #1601 (upper-floor +0.25 m) is HELD — it empirically over-corrects this cohort.
_MAX_SAP_MAE: float = 0.653
# loosening: MAE 0.6517 on the merged tree (with the water-only DHW fix, 16
# Bingley still +6.818). The three silent-drop fixes the gov-API/Elmhurst paths
# already apply; the electric-shower drop (100/205) was the dominant lever.
# Ratcheted 0.653 -> 0.625 by #1602 (floor exposure flags) + #1603 (alternative
# walls + extension floor construction): observed MAE 0.624. #1601 (upper-floor
# +0.25 m) RESOLVED as keep-raw — see the within-0.5 note above.
_MAX_SAP_MAE: float = 0.625
_KNOWN_GAP_REASON = (
"pashub `from_site_notes` mapper does not int-code a main-heating "