ForecastingCrossValidationEvaluationStrategy

EvaluationStrategy

DashAI.back.evaluation.ForecastingCrossValidationEvaluationStrategy

---

A backtest over several validation origins, then the test.

Pair it with RollingOriginSplitter. Each origin fits on the steps before it and is scored on the horizon steps after it; the validation score is the mean over the origins, which is also what a search optimizes. The test is one more origin, at the end.

It runs the same units as the forecasting holdout (KIND = "backtest"). The training steps are never scored, and the kept model is fitted on the whole series (RECONCILING_FORECASTING §2.13).

Methods

get_metadata(cls) -> dict

Defined on BaseEvaluationStrategy

Describe the strategy for the frontend.

Returns

dict

Mapping with kind, which says whether this strategy splits the dataset once or into folds; scored_splits, the partitions it writes metrics for; splitter_partitioning and default_splitter, which splitters it takes and which one a session starts on; splitters, the names it is limited to, or None; and result_views, the views a finished run offers.

Compatible with

• SingleSeriesForecastingTask