Interface: ForecastCertificate

@kortexya/reasoninglayer / Forecast / ForecastCertificate

Interface: ForecastCertificate

Defined in: src/types/forecast.ts:264

What vouches for a committed forecast.

Remarks

The provenance certificate: which model produced the forecast (modelFingerprint, trainWindow), whether the market-odds blend was applied (blended), and the conformal calibration backing any attached covering set or intervals (alpha, coverageSlice, calibrationSize).

Example

console.log(`model ${certificate.modelFingerprint} trained on ${certificate.trainWindow}`);

Properties

alpha?

optional alpha: number

Defined in: src/types/forecast.ts:275

Conformal miscoverage of the attached set/intervals, when present.

blended

blended: boolean

Defined in: src/types/forecast.ts:272

Whether the odds blend was applied (grouped-rank; false otherwise).

calibrationSize?

optional calibrationSize: number

Defined in: src/types/forecast.ts:281

Calibration examples backing the covering set (grouped-rank).

coverageSlice?

optional coverageSlice: string

Defined in: src/types/forecast.ts:278

Mondrian slice the covering set used (grouped-rank).

modelFingerprint

modelFingerprint: string

Defined in: src/types/forecast.ts:266

The producing model’s data fingerprint.

trainWindow

trainWindow: string

Defined in: src/types/forecast.ts:269

The model’s train window.