WalkForwardSplit
Helpers & Utilities ยท AiDotNet.Finance.Trading.Evaluation
Chronological train/holdout split for portfolio experiments: train on the earlier history, evaluate on a strictly-later UNTOUCHED tail, with an optional embargo gap dropped at the boundary so no information leaks across it. This is the honest-eval discipline for time-series RL โ a random split would let the agent train on the future it is later scored on. Mirrors the purged/embargoed split used for supervised training.
Methods
| Method | Summary |
|---|---|
ValueTuple<List<double[]>, List<double[]>> Split(IReadOnlyList<double[]> series, double trainFraction, int embargo) |
Splits each column of series (all equal length) at trainFraction into (train = earlier rows, holdout = later rows), dropping embargo rows at the boundary. |