R/SMART.R

In RMCDA: Multi-Criteria Decision Analysis

#' Apply the SMART Method
#'
#'
#' This function implements the SMART (Simple Multi-Attribute Rating Technique) method in R.
#'
#' @param dataset A numeric matrix or data frame of size (n x m), rows = alternatives, columns = criteria.
#' @param grades A numeric vector of length \code{m} (one grade per criterion).
#'   They get transformed into weights via \eqn{(2^{1/2})^{grades}} and normalized.
#' @param lower A numeric vector of length \code{m} with lower bounds for each criterion.
#' @param upper A numeric vector of length \code{m} with upper bounds for each criterion.
#' @param beneficial.vector A numeric vector containing column indices that are beneficial ("max").
#'
#' @return A matrix (or data frame) named result with two columns: The row index (alternative)
#' and the final SMART score for that alternative.
#'
#' The rows of result are sorted by score in descending order.
#'
#'
#' @examples
#' # Example usage
#' data_mat <- matrix(c(10, 20, 15,  7,
#'                      30,  5,  8, 25),
#'                    nrow = 2, byrow = TRUE)
#' # Suppose we have 4 criteria (2 rows, 4 columns)
#' # We'll treat columns 1, 2, 3 as beneficial, and column 4 as non-beneficial
#' benef_vec <- c(1, 2, 3)
#'
#' # Grades for each of 4 criteria
#' grades <- c(2, 2, 1, 3)
#' lower  <- c(0, 0, 0,  0)
#' upper  <- c(40, 40, 40, 40)
#'
#' # Run SMART
#' result <- apply.SMART(dataset = data_mat,
#'                     grades = grades,
#'                     lower  = lower,
#'                     upper  = upper,
#'                     beneficial.vector = benef_vec)
#'
#' result
#'
#' @export apply.SMART
apply.SMART <- function(dataset,
                        grades,
                        lower,
                        upper,
                        beneficial.vector) {


  X <- as.matrix(dataset)

  #Transform grades into weights

  w <- grades
  for (i in seq_along(w)) {
    w[i] <- (sqrt(2))^w[i]
  }
  w <- w / sum(w)

  n_alt <- nrow(X)
  n_crit <- ncol(X)


  for (i in seq_len(n_crit)) {

    #We'll assume user provides valid bounds; otherwise:
    denom <- upper[i] - lower[i]
    if (denom <= 0) {
      stop(sprintf("Invalid bounds for criterion %d: upper <= lower.", i))
    }


    frac_vec <- (X[, i] - lower[i]) / denom * 64

    #If beneficial
    if (i %in% beneficial.vector) {

      X[, i] <- 4 + log2(frac_vec)
    } else {

      X[, i] <- 10 - log2(frac_vec)
    }
  }


  S <- sweep(X, 2, w, FUN = "*")
  Y <- rowSums(S)


  result <- cbind(index = seq_len(n_alt), score = Y)


  return(result)
}