R/cvts.R

In opoyc/sftools: Automate Statistical Forecasting's workflows

#' Cross-validation for time series data
#'
#' @param ts_data tibble, dataframe or matrix. It is assumed that it is ordered by date.
#' @param cv_size numeric. Cross validation size.
#' @param lag numeric. How many periods ahead to forecast.
#'
#' @return
#'
#' @importFrom purrr map_df
#' @importFrom dplyr tibble
#'
#' @export
#'
#' @examples
#' \dontrun{
#' cvts()
#' }
cvts <- function(ts_data, cv_size, lag){
  # define starting and ending train size
  initial_train <- (nrow(ts_data)-cv_size-lag+1)
  final_train <- initial_train+cv_size-1

  # iterate over to create tupples of train and test
  map_df(initial_train:final_train
         , .f = function(i){
           tibble(train = list(ts_data[1:i,])
                  , test = list(ts_data[i+lag,])
           )}
  )
}