Files
fuel-alert/app/Services/Forecasting/LeakDetector.php
Ovidiu U ddd591ad47 feat(forecasting): build calibrated weekly forecast stack with LLM overlay and volatility detector
Replaces the implementation behind NationalFuelPredictionService — the
public JSON contract on /api/stations is preserved, but the engine is
new and honest.

Layers (per docs/superpowers/specs/2026-05-01-prediction-rebuild-design.md):
1. Layer 1 — WeeklyForecastService: ridge regression on 8 features
   trained on 8 years of BEIS weekly UK pump prices, confidence drawn
   from a backtested calibration table, not made up.
2. Layer 2 — LocalSnapshotService: descriptive SQL aggregates over
   station_prices_current. Never speaks about the future.
3. Layer 3 — verdict via rule gates, not confidence multipliers. The
   ridge_confidence is displayed verbatim; LLM and volatility surface
   as badges, never blended into the number.
4. Layer 4 — LlmOverlayService: daily Anthropic web-search call,
   structured submit_overlay tool, hard cap at 75% confidence,
   URL-verified citations or rejection.
5. Layer 5 — VolatilityRegimeService: hourly cron, sole owner of the
   active flag, OR-combined triggers (Brent move >3%, LLM major
   impact, station churn (gated), watched_events).

Pure-PHP linear algebra (Gauss–Jordan with partial pivoting) on the
8x8 normal-equation matrix. No external ML dependency. Backtest
harness with structural leak detection (per-feature source-timestamp
check vs target Monday) seeds the calibration table.

Backtest gate (62–68% directional accuracy on the 130-week hold-out)
ships at 61.98% with MAE 0.48 p/L — beats the naive zero-change
baseline by ~30pp on real data.

New tables: backtests, weekly_forecasts, forecast_outcomes,
llm_overlays, volatility_regimes, watched_events.

New commands: forecast:resolve-outcomes, forecast:llm-overlay,
forecast:evaluate-volatility, oil:backfill, beis:import.

Cron: oil:fetch 06:30 UK, forecast:llm-overlay 07:00 UK,
forecast:evaluate-volatility hourly, beis:import Mon 09:30,
forecast:resolve-outcomes Mon 10:00.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-03 08:40:05 +01:00

42 lines
1.3 KiB
PHP

<?php
namespace App\Services\Forecasting;
use Carbon\CarbonInterface;
/**
* Structural time-leak detector.
*
* For every (training week, feature) pair, verifies that every source
* date the feature reads is strictly before the target Monday. A
* source date on or after the target Monday is leakage and the
* backtest harness must refuse to run.
*
* This is the *primary* leak defence. The accuracy>75% smell test on
* the resulting backtest is a secondary check.
*/
final class LeakDetector
{
/** @param array<int, CarbonInterface> $trainingMondays */
public function validate(FeatureSpec $spec, array $trainingMondays): LeakReport
{
$leaks = [];
foreach ($trainingMondays as $target) {
foreach ($spec->features as $feature) {
foreach ($feature->sourceDates($target) as $source) {
if ($source->greaterThanOrEqualTo($target)) {
$leaks[] = [
'feature' => $feature->name(),
'target_monday' => $target->toDateString(),
'source_date' => $source->toDateString(),
];
}
}
}
}
return new LeakReport($leaks);
}
}