R/optimize_decay.R

In niallbenj/footballstats: Classify and predict football results

#' @title Optimize Decay
#'
#' @description A function to calculate the weights for a
#'  decaying function.
#'
#' @details This function takes in the number of rows of a
#'  data set being analysed plus a decay factor which is just
#'  an exponent value, then the values of this vector are normalised
#'  to 1 so their weights total 1. (The higher the decay value, the more
#'  weighted to the closest values to the left of the vector become.)
#'
#' @param nrows An integer value defining the number of rows of the
#'  data set being analysed.
#' @param decay A double value that is the exponent value of the decaying
#'  function.
#'
#' @export


optimize_decay <- function(nrows, decay) {

  # Use negative exponential to get decay (severity of decay through `decay`)
  vals <- if (decay > 100) {
    1.0 / nrows %>% rep(nrows)
  } else {
    1:nrows %>%
      `*`(-decay) %>%
      exp()
  }

  # Normalise values to 1 and return
  return(vals %>% `/`(vals %>% sum))
}