Files
fuel-alert/tests/Unit/Services/Forecasting/LeakDetectorTest.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

119 lines
3.5 KiB
PHP

<?php
use App\Services\Forecasting\Contracts\ForecastFeature;
use App\Services\Forecasting\FeatureSpec;
use App\Services\Forecasting\LeakDetector;
use App\Services\Forecasting\LeakReport;
use Carbon\Carbon;
use Carbon\CarbonInterface;
function makeFeature(string $name, array $offsetsInDays): ForecastFeature
{
return new class($name, $offsetsInDays) implements ForecastFeature
{
/** @param array<int, int> $offsetsInDays */
public function __construct(
private readonly string $featureName,
private readonly array $offsetsInDays,
) {}
public function name(): string
{
return $this->featureName;
}
public function valueFor(CarbonInterface $targetMonday): float
{
return 0.0;
}
public function sourceDates(CarbonInterface $targetMonday): array
{
return array_map(
fn (int $offset): CarbonInterface => $targetMonday->copy()->addDays($offset),
$this->offsetsInDays,
);
}
};
}
it('passes when every feature reads strictly before the target Monday', function () {
$spec = new FeatureSpec(
modelLabel: 'test',
features: [
makeFeature('lag_1w', [-7]),
makeFeature('lag_4w', [-7, -14, -21, -28]),
],
);
$report = (new LeakDetector)->validate($spec, [Carbon::parse('2024-06-03')]);
expect($report)->toBeInstanceOf(LeakReport::class)
->and($report->hasLeaks())->toBeFalse()
->and($report->leaks)->toBe([]);
});
it('flags a feature whose source date IS the target Monday', function () {
$spec = new FeatureSpec(
modelLabel: 'test',
features: [makeFeature('same_day', [0])],
);
$report = (new LeakDetector)->validate($spec, [Carbon::parse('2024-06-03')]);
expect($report->hasLeaks())->toBeTrue()
->and($report->leaks)->toHaveCount(1)
->and($report->leaks[0]['feature'])->toBe('same_day');
});
it('flags a feature whose source date is AFTER the target Monday', function () {
$spec = new FeatureSpec(
modelLabel: 'test',
features: [makeFeature('future', [7])],
);
$report = (new LeakDetector)->validate($spec, [Carbon::parse('2024-06-03')]);
expect($report->hasLeaks())->toBeTrue()
->and($report->leaks[0]['feature'])->toBe('future')
->and($report->leaks[0]['target_monday'])->toBe('2024-06-03')
->and($report->leaks[0]['source_date'])->toBe('2024-06-10');
});
it('checks every training week, not just the first', function () {
$spec = new FeatureSpec(
modelLabel: 'test',
features: [makeFeature('lag_1w', [-7])],
);
$weeks = [
Carbon::parse('2024-06-03'),
Carbon::parse('2024-06-10'),
Carbon::parse('2024-06-17'),
];
$report = (new LeakDetector)->validate($spec, $weeks);
expect($report->hasLeaks())->toBeFalse();
});
it('reports multiple leaks across multiple features', function () {
$spec = new FeatureSpec(
modelLabel: 'test',
features: [
makeFeature('clean', [-7]),
makeFeature('leaky_one', [0]),
makeFeature('leaky_two', [3]),
],
);
$report = (new LeakDetector)->validate($spec, [Carbon::parse('2024-06-03')]);
expect($report->hasLeaks())->toBeTrue()
->and($report->leaks)->toHaveCount(2);
$featureNames = array_column($report->leaks, 'feature');
expect($featureNames)->toContain('leaky_one', 'leaky_two')
->and($featureNames)->not->toContain('clean');
});