Build a geographically DENSE postcode-clustered corpus to test cross-postcode
geo expansion (the handover's anticipated "real geo payoff"). The gov EPC API
has no area/prefix search (a partial postcode 400s; the old opendatacommunities
partial-search API is decommissioned), so neighbourhood enumeration is external:
seed K postcodes nationally, expand each via postcodes.io's nearest-postcode
endpoint into every unit within RADIUS_M, pull each one's full EPC cohort.
postcodes.io is a corpus-BUILD dependency only — the predictor stays pure. Same
on-disk layout as the scattered corpus, so load_corpus + the coords resolver
consume it unchanged.
MEASURE-FIRST RESULT — cross-postcode expansion is a NO-GO. On a 2-seed pilot
(York YO19 + Islington N51, 81 postcodes / 1558 certs, 140 SAP-10.2 targets),
pooling nearby postcodes regresses accuracy across the board:
same-postcode FA_MAE 9.53 wall 92% age 72% floor_con 85% cylinder 91%
cross <=0.3km FA_MAE 13.1 wall 80% age 61% floor_con 82% cylinder 79%
Even as a thin-cohort top-up it hurts (thin n=18: FA 5.24 -> 7.15). Root cause:
the postcode boundary is itself a strong homogeneity prior (a postcode is one
coherent street/development), so same-postcode neighbours beat geographically
near cross-boundary ones even when the home postcode is sparse (and they rarely
are — median same-postcode cohort here is 34). Geo-proximity helps WITHIN a
postcode (#1227) but does not survive crossing the boundary. Cross-postcode geo
closed; geo weighting stays intra-postcode. Tooling kept (reusable).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Size the predicted dwelling from the geo-proximity-weighted median of the
cohort's floor areas rather than the plain median: homes built together share a
footprint, so a nearer neighbour's area should count for more (the same street
signal #1227 already wired into age / wall / glazing). Reuses `_geo_weights` and
adds `_weighted_median`, which reduces exactly to `statistics.median` under
uniform weights (geo off / no target coordinates) — including the even-count
midpoint average — so the MAD-minimising guarantee is preserved.
Measured over the 514-target SAP-10.2 corpus (leave-one-out):
floor_area MAE 10.48 -> 9.73 m² MAPE 13.2% -> 12.2%
Re-baselines the n=36 fixture floor_area ceiling 11.8983 -> 12.0378 (a method
change, not a loosening; the small fixture subset moved +0.14 the other way as
sample noise while the population improved decisively). The ceiling still pins
the new deterministic value exactly, so the tighten-only ratchet resumes.
Investigation ruling out the adjacent floor-area levers (kept in the follow-up):
lowering minimum_cohort (9.78-10.03, worse), hard same-form filter (10.19),
mean instead of median (10.68), constant bias correction (10.47),
extension-conditioning (oracle 9.50, not worth the misclassification cost) and
room-in-roof conditioning/additive (RiR is a confound for large multi-part
outliers — RiR area is only ~21% of total, and the increment breaks the homes
already predicted exactly). Remaining cohort lever is built-form soft-weighting,
gated on a denser corpus.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds a floor_area line giving MAE (m2), MAPE (% of actual), and the typical
(median actual) size, so the absolute error reads relative to dwelling size.
Corpus: MAE 10.48 m2 / MAPE 13.2% / typical 61 m2.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
main's 'ND' multiple_glazing_type mapper fix (361abc12) changes the mapped
ground-truth glazing for one fixture cert, so modal_glazing_type re-baselines
0.5833 -> 0.5556 (21/36 -> 20/36). A mapper change shifts the deterministic
fixture rates like a fixture change does — re-baseline, not a prediction
regression. All other component floors + residual ceilings unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Folds a haversine distance kernel into the categorical-mode weighting so a
nearer neighbour counts for more — applied ONLY to the components that showed
a clear distance signal in the corpus pre-check (age band, wall + floor
construction, glazing: homes built/retrofitted together cluster). Roof
construction showed no decay and is excluded; heating keeps its coherent
donor. Predictor stays pure: weights come from target.coordinates vs each
Comparable.coordinates (resolved at the boundary); geo is OFF when the target
has no coords, neutral for a neighbour with none.
Scale chosen on the harness: _GEO_SCALE_KM=0.1 is the gate-safe optimum
(0.05 lifts the corpus more but regresses fixture floor_construction).
Corpus (150pc/514, geo off->on): age 0.564->0.572, age_pm1 0.841->0.847,
wall 0.902->0.912, floor_con 0.786->0.796, glazing 0.667->0.673; roof
unchanged. Fixture: glazing 0.5278->0.5833 (floor ratcheted), all else held.
Refactored recency into a reusable _recency_weights vector composed via
_combine, so similarity/recency/geo factors multiply uniformly. Fixture ships
a committed _coordinates.json (OGL OS OpenData; build script carries it from
the corpus sidecar on rebuild) so the gate exercises geo without S3.
This is the per-component method applied to geography ([[feedback_per_component_best_method]]).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds coordinates: Optional[Coordinates] to Comparable and PredictionTarget
(data carriers — the pure predictor stays IO-free), and wires load_corpus to
read an optional _coordinates.json sidecar ({uprn: [lon, lat]}) and populate
each Comparable from its cert's uprn; iter_predictions threads the held-out
target's coordinates through. Absent sidecar -> geo-weighting stays off (no
behaviour change yet — weighting lands next slice). fetch_corpus_coordinates
now writes the sidecar into the corpus dir. load_corpus populates 99% of
corpus comparables.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds GeospatialRepository.coordinates_for_uprns(uprns) -> dict — a batch
coordinate lookup returning only covered UPRNs. The S3 adapter overrides it
to read the meta once, group UPRNs by their covering partition, and read each
partition once for all the UPRNs it covers; co-located (closely-numbered)
UPRNs share a partition, so an EPC Prediction cohort is typically one or two
reads instead of one per neighbour. Default port impl is a per-UPRN loop.
Feeds the EPC Prediction geo-proximity work: a cohort's UPRNs resolve to
coordinates in a couple of reads (validated at corpus scale: 170 partition
reads for 2683 UPRNs).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
One-time utility: resolves every corpus cert's uprn -> WGS84 lon/lat from the
OS Open-UPRN parquet (DATA_BUCKET/spatial/) via boto3, grouping UPRNs by their
covering partition so each ~1.7MB partition is read at most once (the efficient
batch lookup we intend to add to GeospatialRepository). Caches {uprn:[lon,lat]}
locally for the validation harness. Resolved 2609/2683 corpus UPRNs (97%).
Signal pre-check result (does intra-postcode proximity predict components?):
intra-postcode distances are non-trivial (median 44m, p90 138m, max ~1km),
and nearer neighbours match the target markedly better on age band (0.63 at
<20m -> 0.16 at >300m), wall, glazing and floor construction. Roof shows no
decay. => geo-proximity is worth building, per-component (strongest for age,
the weakest fabric component).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds roof_insulation_thickness_pm1 (mirrors construction_age_band_pm1, issue
#1222): adjacent RdSAP thickness buckets (0/NI,12mm..400mm+) carry near-
identical roof U-values, so an off-by-one bucket is a SAP-neutral hit. 'ND'
(no-data) is off the ordered scale, so only an exact match counts there.
Honest measurement of SAP-relevant roof-insulation quality.
Corpus (150pc/514): exact 49.3% -> +/-1 53.7% (the misses are often multiple
buckets or ND, so the band gain is smaller than age's). Fixture: exact ==
+/-1 (0.4118) — its misses are all >1 bucket; gate floor added at 0.4118.
Also fixes two pre-existing pyright errors in the touched test file
(_epc main_fuel_type/main_heating_control were Optional but the
MainHeatingDetail attributes are non-optional Union[int, str]).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Three more pre-existing failures (present at 9ee38211, before this branch's
recent commits; same family as the orchestration multi-measure re-pin) —
golden-cert plan expectations that predate the ASHP generator (ADR-0025)
and the optimiser folding forced dependencies into candidate gain (ADR-0016):
- test_console: a multi-measure plan now leads with air_source_heat_pump,
not cavity_wall_insulation (which is dropped — its forced ventilation makes
the pair net-negative). Assert a measure actually in the package.
- test_report 0330: package is now {solid_floor_insulation, air_source_heat_
pump}; cavity_wall + forced mechanical_ventilation correctly excluded.
- test_report 0036: gain-maximising package is now {solid_floor_insulation,
low_energy_lighting}.
Same verified-correct optimiser evolution as 077e3a39 (cavity_wall +2.9 SAP
alone but its forced fabric→ventilation dep drags the pair net-negative).
Re-pin to the actual packages + their trigger fields; the forced wall→vent
edge stays covered by test_measure_dependency / test_optimiser.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The validation report showed only the SAP calculator floor (calc(actual) vs
lodged), so the headline PEI MAE (~40 kWh/m2) read as prediction error when
much of it is the calculator's own API-path residual. Adds the CO2 + PEI
floors alongside SAP.
Diagnostic (150pc/514): PEI floor MAE 15.73 (calc(actual) vs lodged) vs SAP
floor 1.57; calc(actual)/lodged PEI ratio ~1.06 (mean +10.7, ~+6% over-
estimate). That RULES OUT the suspected gross unit/definition mismatch (a
unit bug would be ~2x/3.6x, not 1.06) and reframes #1228: the PEI gap is a
modest calculator bias (~16 floor, calc-branch) plus a larger prediction-
sensitivity term (~24) — PEI is far more prediction-sensitive than SAP.
CO2 floor 0.20 t. Script-only; no gate impact.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Heating sub-fields can't be field-moded without breaking system coherence,
so the whole SapHeating cluster is now copied as a unit from a single
coherent donor rather than inherited from the structural template: the
neighbour matching the cohort's modal heating signature (main fuel +
category + cylinder presence), most recent among the matches (recent cert =
current system). Including cylinder presence in the signature is load-bearing
— it protects has_hot_water_cylinder + cylinder_insulation (a bare fuel+cat
signature regressed them).
Corpus (150pc/514): heating_main_control 66.3 -> 73.9% (+7.6, the target),
main_fuel 92.8 -> 96.9, category 90.7 -> 95.7, water_fuel 92.8 -> 96.3,
water_code 88.5 -> 95.3, has_cylinder 81.1 -> 89.7, secondary 36.2 -> 42.0.
SAP MAE vs lodged 7.08 -> 6.00 (calculator floor 1.57). cylinder_insulation
-13.6 corpus (tiny-n) but +33pp on the fixture; AC requires control up +
fuel/category hold + SAP not worsened, all met.
Gate (36-target fixture): zero regression; ratcheted main_category
0.8889->0.9444, main_control 0.7500->0.8056, water_fuel 0.9167->0.9722,
water_code 0.8889->0.9444, cylinder_insulation_type 0.1667->0.5000. This is
the per-component heating method ([[feedback_per_component_best_method]]):
coherent donor, never field-mode.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The optimiser-package expectation was stale: it predated the optimiser
folding a triggered measure's forced dependency into its candidate gain
(ADR-0016). The run considers ALL measures (considered_measures defaults
to None — no restriction), so once the ASHP bundle became SAP-beneficial
(ADR-0025) the gain-maximising package shifted.
Verified the new package is CORRECT, not a regression: on the test EPC,
cavity-wall insulation earns +2.9 SAP alone but its forced fabric→
ventilation dependency (ADR-0016) drags the wall+ventilation pair to a
NET −1.8 SAP (−0.9 on top of the ASHP package), so the gain-maximising
Optimiser correctly excludes the wall and its forced ventilation. Update
the expected set to {air_source_heat_pump, suspended_floor_insulation,
low_energy_lighting, secondary_heating_removal} and drop the wall/vent-
specific assertions — the forced wall→ventilation edge is covered by
test_measure_dependency / test_optimiser; this integration test keeps its
end-to-end optimise→persist→telescope coverage on the chosen package.
Pre-existing failure (present before this branch's recent commits), outside
the handover regression gate.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Per-component method: glazing type is now the recency-weighted cohort mode
applied to every predicted window, rather than copied from the template.
Glazing is retrofitted over a dwelling's life (single -> double), so a
recent neighbour reflects the current state — same family as roof-insulation
thickness. Recency is the CORRECT weighting here: plain moding regressed the
fixture (-5.6pp) and was previously reverted; similarity weighting also
regressed it; recency improves BOTH (window geometry stays on the template,
only the glazing categorical moves).
modal_glazing_type: corpus (150pc/514) 60.7 -> 66.7% (+6.0pp); fixture
0.5000 -> 0.5278 (floor ratcheted up). Heating, geometry residuals and all
other components unchanged. Refactored _recency_weighted_mode to a reusable
_recency_weighted_choice(value_of) shared by roof insulation + glazing.
Closes the #1223 per-component approach: floor-area (median estimate) +
glazing (recency) shipped as distinct best-fit methods rather than a global
recency template, which would have disturbed the coherence-coupled heating
cluster.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The realistic re-generation of case 37 (code-117 gas boiler, control 2102,
+ a MAINS-GAS condensing gas-fire secondary code 611, vs case 37's biogas
605). The full extractor -> mapper -> calculator pipeline reproduces the
worksheet's SAP-rating block EXACTLY: continuous SAP 60.9152 (Δ 2e-5) and
(272) CO2 5801.0770 (Δ ~0). This confirms the boiler-efficiency /
control-2102 −5pp interlock / secondary-fuel handling are all correct, and
that case 37's +7 gap was purely the biogas sub-fuel the Summary export
cannot carry.
Summary mirrored into backend/documents_parser/tests/fixtures so the pin
runs without the unstaged workspace. PE not pinned — it is a separate
DPER block (different scope) already guarded by the corpus PE gauge.
Worksheet harness 47/47 unchanged; pyright net-zero.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Per-component method, not a global template change: the predicted floor
area is now the cohort median (the MAD-minimising point estimate of the
target's size) rather than whichever structural template's own area. The
calculator derives heat loss from building-part geometry, not this scalar,
so decoupling them is safe and the scalar becomes a better size estimate.
floor_area mean|.|: corpus (150pc/514 targets) 10.62 -> 10.48; fixture
12.2175 -> 11.8983 (ceiling ratcheted down). No other component moves.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The HHR-storage HeatingOverlay (ADR-0024) added an off-peak electric
immersion cylinder but never set `immersion_heating_type`, so the overlaid
cert left it None. The calculator then could not resolve `immersion_single`
for the SAP 10.2 Table 13 HW high-rate split and billed hot water 100% at
the off-peak low rate — £127.41 vs the relodged after-cert's £169.39,
overstating the overlay's SAP by +1.26 (CO2/PE matched, isolating it to the
HW cost path).
Add `immersion_heating_type` to HeatingOverlay, route it through
`_fold_heating` (it lives on `sap_heating`), and set it to 1 (single
off-peak immersion) on the HHR overlay to match the relodged reference.
Closes both `test_hhr_storage_overlay_reproduces_the_relodged_after_*`
cascade pins (electric-storage and no-system befores share the after).
Pre-existing failure (present before this branch's recent commits), outside
the handover regression gate. Full modelling suite 220 pass, pyright net-
zero.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
`_synthesise_20_0_0_sap_windows` passed `schema.multiple_glazing_type`
straight into `_api_cascade_glazing_type`, which raised UnmappedApiCode on
the "ND" (Not Defined) string that the 20.0.0 corpus lodges alongside the
1-8 integer codes — failing the mapper-coverage guard on every ND-glazed
20.0.0 cert. Mirror the existing 18.0/19.0/17.x seams: route integer codes
through the cascade, fall the "ND" string back to the DG-modal default
(cascade code 2 → daylight g_L 0.80). Also corrects the 20.0.0 schema
field type `int` → `Union[int, str]` to match the data (as 18.0 already
does), which keeps the isinstance guard pyright-clean.
Pre-existing failure (present before this branch's recent commits), not in
the handover regression gate. Fixes all 15 RdSAP-Schema-20.0.0 ND certs;
test_mapper_corpus 6002/6002 pass. pyright net-zero.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Bumps N_POSTCODES 40 -> 150 for the fetch script. Larger corpus (150
postcodes / 3719 certs) reduces leave-one-out variance and unblocks the
recency-template work (#1223), which regressed the noisier 36-target gate
fixture. Corpus itself stays out of git (gitignored /tmp + persistent
backup at /workspaces/home/epc_prediction_corpus_backup).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Completes `_elmhurst_secondary_fuel_from_sap_code` per SAP 10.2 §12
(PDF p.34: "Secondary heating systems and applicable fuel types are taken
from the room heaters section of Table 4a") + RdSAP 10 §10.4.1. Each
Table 4a room-heater code now resolves to its fuel CATEGORY's modal fuel:
- gas room heaters 601-613 → mains gas (26 → Table 32 1, 3.48 p/kWh)
- liquid room heaters 621-625 → heating oil (28 → Table 32 4, 5.44 p/kWh)
- solid room heaters 631-636 → house coal (11 → Table 32 11, 3.67 p/kWh)
- electric room htrs 691-694/699/701 → None (cascade electricity default)
Previously only the gas (601-613→26) and solid (631-634→11) blocks were
mapped; liquid heaters (621-625) and 635-636 fell through to None →
silently billed as electricity (13.19 p/kWh), a large mis-price for an
oil/solid heater. The prior slice raised on those; this maps them to the
correct category fuel instead, and keeps the raise ONLY for codes inside
the room-heater range (601-701) that are not a recognised Table 4a row.
The specific sub-fuel within a category (mains gas vs LPG vs biogas) is a
SEPARATE lodgement per §10.4.1 and is NOT exported in the Summary, so the
gas block stays the modal mains gas — worksheet "simulated case 37" lodged
its 605 live-effect fire on biogas (7.60 p/kWh), unrecoverable from the
Summary code alone (this is the entire +7 SAP case-37 gap: secondary
energy £131 + a separate biogas standing charge £70; every other line
matches the worksheet exactly, incl. (206) main efficiency 61%).
5 AAA tests, harness 47/47 (0 raised), pyright net-zero, regression clean,
corpus gauge unchanged (Elmhurst-path only).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
ADR-0029 decision 5: survivors were treated equally; now each neighbour's
vote in the cohort mode decays with its distance from the cohort's physical
centre (floor area from the median, age band from the modal band), so the
mode leans on the most representative neighbours instead of being swayed by
size/era outliers. Scales (size 20 m^2, age weight 0.5) chosen on the
validation corpus; the tight size kernel is load-bearing (looser scales
regress floor_insulation on the fixture).
Corpus (181 SAP-10.2 targets): wall_insulation 83.4->86.2%,
roof_construction 86.2->87.3%, floor_construction 78.8->81.2%,
floor_insulation 92.9->94.1%; net +7.5pp gained vs -1.1pp (two 1-cert dips,
both held on the fixture). Geometry/residuals untouched (template unchanged).
Gate (36-target fixture): zero regression across all 24 floors/ceilings;
ratcheted wall_insulation_type 0.7778->0.8333, floor_construction
0.7500->0.8125, floor_insulation 0.9062->0.9375. Dead _mode/_int_mode
removed (superseded by the weighted variants).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds PredictionConfidence (cohort size + per-component agreement = the
modal value's share among neighbours that lodge one) and
EpcPrediction.confidence(), a compute-only signal so downstream can flag
low-confidence components (ADR-0029 open item: 'confidence signal').
Sanity check on the 40-postcode corpus (1068 component predictions):
agreement is strongly predictive of correctness — pooled hit-rate 21.9%
(<0.5) / 46.7% (0.5-0.7) / 73.6% (0.7-0.9) / 95.5% (>=0.9); point-biserial
corr(agreement, correct) = 0.582. Cohort size tracks too (<6 -> 68.4%,
>=20 -> 96.0%). Surfacing / persistence is a separate HITL follow-up.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The Elmhurst Summary lodges only the secondary heating SAP code (Table 4a
Category 10), never its fuel. `_elmhurst_secondary_fuel_from_sap_code`
mapped the gas block (601-613 → mains gas) and solid block (631-634 →
house coal) to their modal defaults, but returned None for any OTHER
Category-10 code — and None makes the cascade SILENTLY bill the secondary
as electricity (13.19 p/kWh). For a fuel-fired heater (e.g. 621-625
liquid-fuel oil/bioethanol) that is a large, invisible mis-price.
Per the UnmappedElmhurstLabel strict-raise pattern (mirrors the wall_type
/ glazing label raises), a fuel-fired Category-10 code (601-699) outside
the mapped gas/solid blocks now RAISES instead of guessing. Electric room
heaters (691-699) keep returning None — electricity IS their fuel.
The gas block 601-613 still resolves to the modal default mains gas: the
Summary cannot distinguish mains gas from LPG/biogas, so an LPG or biogas
live-effect fire (worksheet "simulated case 37" used biogas at 7.60 p/kWh
vs our 3.48 p/kWh mains-gas default, a +7 SAP gap) is not recoverable from
the Summary export — that is a data-availability limit, not a guess we can
fix here. This commit closes the genuinely-silent-wrong path; the gas
sub-fuel remains the documented modal default.
Worksheet harness 47/47, 0 raised. 3 AAA tests, pyright net-zero,
regression clean, corpus gauge unchanged (Elmhurst-path only; the API path
lodges the secondary fuel explicitly).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Measurement honesty so we optimise SAP-relevant accuracy, not SAP-neutral
misses (ADR-0030 Component Accuracy):
- Add construction_age_band_pm1: an exact-or-adjacent-band hit. Adjacent
RdSAP age bands carry near-identical U-values, so an off-by-one is
~SAP-neutral. Full corpus: exact 78.5% but ±1-band 91.7% (fixture
63.9% -> 83.3%) — most age misses are adjacent.
- Drop window_count from the gate's residual ceilings (cosmetic): the
predicted picture clusters at a mapper-default 4 windows vs actuals 1-21,
but total_window_area (the SAP-relevant signal) stays tight at ~3.4 m2.
Gate: + construction_age_band_pm1 floor 0.8333; window_count no longer gated.
Closes#1222
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
End-to-end API-path regression pin for the battery behaviour validated by
the user-simulated Elmhurst worksheet pair (cert 001431 "simulated case
35/36", 5 kWh, export-capable, mains-gas, standard tariff). The official
SAP rating ("10a. Fuel costs - using Table 12 prices") values PV used-in-
dwelling and PV exported identically at 13.19 p/kWh (export code 60 ==
import code 30, ADR-0010), so a battery only redistributes PV between two
equally-priced lines: worksheet PV credit (252) = -455.6458 and SAP (258)
= 88.0859 are IDENTICAL with/without the battery (ΔSAP = 0).
Two tests over the committed RdSAP-21.0.1 corpus:
- standard tariff (meter 2): toggling the battery holds continuous SAP
EXACTLY constant, while at least one cert's primary energy DOES respond
(proving the App-M1 §3c β-split is wired, not a dropped battery).
- off-peak tariff (meter != 2): the battery STRICTLY raises SAP, because
self-consumed PV displaces high-rate import (15.29) above the 13.19
export credit — confirming the standard-tariff neutrality is a price
coincidence, not a no-op.
Guards table_32 export price (code 60) and the battery β-split against
silent regression. Complements the unit-level β tests in
test_photovoltaic.py.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Investigated recency-weighting (weight cohort votes by an exponential decay
in cert age). Key finding: it must be SELECTIVE. On the validation corpus it
HURTS permanent categoricals (wall 91.2->89.5, age 78.5->75.7 — discards
still-valid data) but clearly HELPS time-varying ones, where a recent
neighbour reflects the current physical state:
roof_insulation_thickness 56.7 -> 60.7% corpus (+4pp)
29.4 -> 41.2% fixture (+12pp)
So apply a recency-weighted mode only to roof_insulation_thickness (loft
top-ups happen over time); keep the plain mode for permanent categoricals.
tau = 4yr (~2.8yr half-life); falls back to plain mode when no registration
dates are lodged. Gate floor ratcheted 0.2941 -> 0.4118.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
These independent fabric categoricals were template-copied; mode them like
the construction categoricals. Verified mode beats template before applying.
Big fixture win on roof insulation thickness (doubled), floor insulation
neutral-to-positive:
roof_insulation_thickness 14.7% -> 29.4% (gate floor ratcheted up)
floor_insulation 90.6% (unchanged on the fixture)
Glazing type was tried too (+1.6pp on the 40-postcode corpus) but REGRESSED
the 36-target fixture (0.50 -> 0.44) — the gate caught it. Glazing moding is
marginal/noisy, so it's left on the template; revisit with a larger corpus.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Tier-2 (full national bulk streaming) is deferred. The near-term scale
validation is a Tier-1.5: a few-thousand-cert anonymised corpus stored in
S3 (too large to commit, far more stable than the 36-target gate fixture),
pulled to a temp dir and run through the same load_corpus +
evaluate_component_accuracy. Reuses the committed-fixture machinery wholesale
— only the data source differs. One scorer, three data sources (committed
fixture / S3 corpus / bulk stream).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The committed CI gate: run the calculator-free leave-one-out scorer over the
frozen anonymised fixture (36 SAP-10.2 targets) and assert each per-component
classification rate / geometry residual is no worse than a committed baseline.
Prediction is deterministic + the fixture frozen, so the numbers reproduce
exactly — a failure is a real regression, never sample noise.
- 19 rate floors + 5 residual ceilings, seeded at the currently-measured
values; they only ever tighten (no-widening ethos on an aggregate).
- Calculator-FREE — component floors are the real gate; the end-to-end
SAP/carbon/PE guards stay out (their floor is the separate API-path
calculator workstream).
- Skips with a message when the fixture is absent.
25 parametrized assertions, all green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
