CV methods

timecave.validation_methods.CV

This module contains all the CV ('Cross-Validation') validation methods supported by this package.

Classes:

Name	Description
BlockCV	Implements the Block CV method, along with its weighted version.
hvBlockCV	Implements the hv Block method.
AdaptedhvBlockCV	Implements the Adapted hv Block CV method, along with its weighted version.

See also

Out-of-Sample methods: Out-of-sample methods for time series data.

Prequential methods: Prequential or forward validation methods for time series data.

Markov methods: Markov cross-validation method for time series data.

Notes

Cross-validation methods are one of the three main classes of validation methods for time series data (the others being out-of-sample methods and prequential methods). Like prequential methods, CV methods partition the series into equally sized folds (with the exception of the hv Block variant). However, CV methods do not preserve the temporal order of observations, meaning that a model can be trained on later data and tested on earlier data. CV methods also differ from Out-of-Sample methods, as the latter do not partition the series in the same way. For more details on this class of methods, the reader should refer to [1].

References

1

Vitor Cerqueira, Luis Torgo, and Igor Mozetiˇc. Evaluating time series forecasting models: An empirical study on performance estimation methods. Machine Learning, 109(11):1997–2028, 2020.