validation_split: Validation Split

In yvesmauron/univariate-time-series-forecasting: Intelligent Algorithms to be Integrated in Business Intelligence solutions.

Description

This function splits a time series into validation sets with training and test data; the output can be used to backtest models and/or select/combine models based on how they perfrom on this sets.

Usage

validation_split(
  y,
  val_h,
  sov_only = F,
  val_min_years = 4,
  cv_min_years = 5,
  cv_max_samples = 3,
  ...
)

Arguments

y	Time series onject.
val_h	Horizon to be used for validation.
sov_only	Flag, whether only single-origin validation should be considered. Optional, default TRUE.
val_min_years	Minimum years required to conduct single origin validation. Optional, default 4.
cv_min_years	Minimum years required to conduct cross-origin validation. Optional, default 5.
cv_max_samples	Maximum samples that should be considered during cross-validation, i.e. 3 indicates the algorithm validates from 3 origins. Optional, default 3.
...	not used

Value

list of validation sets

yvesmauron/univariate-time-series-forecasting documentation built on March 2, 2020, 12:20 a.m.