R/LoB.Backtest.R

In EduardoRamosP/MackNet: Mack-Net model: Blending Mack's model with Recurrent Neural Networks

#' @title LoB.Backtest
#' @description It computes the percentile of the observed data vs the distribution generated by AY and totals.
#' @param Reservest.D Total reserves sampled by the model.
#' @param Reservest.D.Vector Total reserves by AY sampled by the model.
#' @param Reserves.Validation Actual reserve.
#' @return Percentile of the observed data vs the distribution generated by AY and totals.
#' @import keras
#' @import abind
#' @export
#'

LoB.Backtest=function(Reservest.D,Reservest.D.Vector,Reserves.Validation){
  dimension=dim(Reservest.D.Vector)[1];Out=0
  for (i in 2:dimension){
    Dist=ecdf(Reservest.D.Vector[i,])
    Out[i]=Dist(Reserves.Validation[i])
  }
  Dist=ecdf(Reservest.D)
  Out[i+1]=Dist(sum(Reserves.Validation))
  return(Out)
}