PurgedWalkForwardCrossValidator<T, TInput, TOutput>

Models & Types · AiDotNet.CrossValidators

Walk-forward cross-validation that purges training samples whose label windows overlap the validation fold, and embargoes a gap after it.

Inheritance: CrossValidatorBase<T, TInput, TOutput> → PurgedWalkForwardCrossValidator<T, TInput, TOutput>

For Beginners

When the answer for “today” depends on the next 20 days, the rows just before your validation period secretly contain validation-period information. Scoring against them flatters the model. This removes them, so the number you get is one you can trust.

How It Works

Every other cross-validator here — including TimeSeriesCrossValidator, which is rolling-origin — leaks when labels look forward. A 20-bar forward return computed at the last training index is derived from bars inside the validation fold, so the model is scored partly on data it effectively saw, and the resulting estimate is optimistic. Purging removes those overlapping training samples; the embargo additionally drops a gap after the fold to cover serial correlation outliving the label window.

The fold geometry is PurgedWalkForwardValidator’s — the same logic behind PurgedWalkForwardSplitter, so a purged split and a purged cross-validation agree rather than being two separate implementations that can drift apart.

Constructors

Constructor Summary
PurgedWalkForwardCrossValidator(int labelHorizon, int nSplits, int embargo, bool expanding, CrossValidationOptions options) Creates a purged, embargoed walk-forward cross-validator.

Methods

Method Summary
CrossValidationResult<T, TInput, TOutput> Validate(IFullModel<T, TInput, TOutput> model, TInput X, TOutput y, IOptimizer<T, TInput, TOutput> optimizer)