R/wc_recalculate.R

In whiboclustering: White Box Clustering Algorithm Design

#' General Component for Cluster Representative update.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param recalculate_type String which signal which update type to be used. Check \code{wc_recalculate_types} for possible values.
#' @param assignment_type Assignment type (Optional).
#' @param old_centroids Old centroids (Optional).
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalculate <- function(data, assignment, recalculate_type, assignment_type = NULL, old_centroids = NULL)
{
  # CHECKING FOR ERRORS
  if (!(tolower(recalculate_type) %in% tolower(wc_recalculate_types$Type)))
  {
    stop('Please enter assignment function that is available in wc_assign_types data frame')
  }

  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  #RECALCULATE CLUSTER REPRESENTATIVE
  if(grepl(pattern = 'online', x = recalculate_type, ignore.case = TRUE))
  {
    centroids <- eval(call(name = as.character(wc_recalculate_types$Method[tolower(wc_recalculate_types$Type) == tolower(recalculate_type)]), data, assignment, assignment_type, old_centroids))
  }
  else
  {
    centroids <- eval(call(name = as.character(wc_recalculate_types$Method[tolower(wc_recalculate_types$Type) == tolower(recalculate_type)]), data, assignment, old_centroids))
  }

  return(centroids)
}

#' Solution for Cluster Representative update which uses Mean.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_mean <- function(data, assignment, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  new_centroids <- stats::aggregate(x = data, by = list(assignment), FUN = mean)
  colnames(new_centroids)[1] <- "WCCluster"

  #If some cluster does not have assignments, add old one
  if(!is.null(old_centroids))
  {
    if(nrow(new_centroids) != nrow(old_centroids))
    {
      new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster %in% setdiff(old_centroids$WCCluster, new_centroids$WCCluster),])
    }
  }

  return(new_centroids)
}

#' Solution for Cluster Representative update which uses Median.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_median <- function(data, assignment, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  new_centroids <- stats::aggregate(x = data, by = list(assignment), FUN = stats::median)
  colnames(new_centroids)[1] <- "WCCluster"

  #If some cluster does not have assignments, add old one
  if(!is.null(old_centroids))
  {
    if(nrow(new_centroids) != nrow(old_centroids))
    {
      new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster %in% setdiff(old_centroids$WCCluster, new_centroids$WCCluster),])
    }
  }

  return(new_centroids)
}

#' Solution for Cluster Representative update which uses Geometric mean.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_geometric_mean <- function(data, assignment, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  new_centroids <- stats::aggregate(x = data, by = list(assignment), FUN = function(x) {prod(x)^(1/length(x))})
  colnames(new_centroids)[1] <- "WCCluster"

  #If some cluster does not have assignments, add old one
  if(!is.null(old_centroids))
  {
    if(nrow(new_centroids) != nrow(old_centroids))
    {
      new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster %in% setdiff(old_centroids$WCCluster, new_centroids$WCCluster),])
    }
  }

  return(new_centroids)
}

#' Solution for Cluster Representative update which uses Harmonic mean.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_harmonic_mean <- function(data, assignment, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  new_centroids <- stats::aggregate(x = data, by = list(assignment), FUN = function(x) {length(x)/(sum(1/x))})
  colnames(new_centroids)[1] <- "WCCluster"

  #If some cluster does not have assignments, add old one
  if(!is.null(old_centroids))
  {
    if(nrow(new_centroids) != nrow(old_centroids))
    {
      new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster %in% setdiff(old_centroids$WCCluster, new_centroids$WCCluster),])
    }
  }

  return(new_centroids)
}

#' Solution for Cluster Representative update which uses Quadratic mean.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_quadratic_mean <- function(data, assignment, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  new_centroids <- stats::aggregate(x = data, by = list(assignment), FUN = function(x) {sqrt(1/length(x) * sum(x^2))})
  colnames(new_centroids)[1] <- "WCCluster"

  #If some cluster does not have assignments, add old one
  if(!is.null(old_centroids))
  {
    if(nrow(new_centroids) != nrow(old_centroids))
    {
      new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster %in% setdiff(old_centroids$WCCluster, new_centroids$WCCluster),])
    }
  }

  return(new_centroids)
}

#' Solution for Cluster Representative update which uses Trimmed mean.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_trimmed_mean <- function(data, assignment, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  new_centroids <- stats::aggregate(x = data, by = list(assignment), FUN = mean, trim = 0.05)
  colnames(new_centroids)[1] <- "WCCluster"

  #If some cluster does not have assignments, add old one
  if(!is.null(old_centroids))
  {
    if(nrow(new_centroids) != nrow(old_centroids))
    {
      new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster %in% setdiff(old_centroids$WCCluster, new_centroids$WCCluster),])
    }
  }

  return(new_centroids)
}

#' Solution for Cluster Representative update which uses Trimean.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_trimean <- function(data, assignment, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  new_centroids <- stats::aggregate(x = data, by = list(assignment), FUN = function(x) {0.5 * (stats::quantile(x)[[3]] + (stats::quantile(x)[[2]] + stats::quantile(x)[[4]])/2)})
  colnames(new_centroids)[1] <- "WCCluster"

  #If some cluster does not have assignments, add old one
  if(!is.null(old_centroids))
  {
    if(nrow(new_centroids) != nrow(old_centroids))
    {
      new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster %in% setdiff(old_centroids$WCCluster, new_centroids$WCCluster),])
    }
  }

  return(new_centroids)
}

#' Solution for Cluster Representative update which uses Midhinge.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_midhinge <- function(data, assignment, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  new_centroids <- stats::aggregate(x = data, by = list(assignment), FUN = function(x) {(stats::quantile(x)[[2]] + stats::quantile(x)[[4]])/2})
  colnames(new_centroids)[1] <- "WCCluster"

  #If some cluster does not have assignments, add old one
  if(!is.null(old_centroids))
  {
    if(nrow(new_centroids) != nrow(old_centroids))
    {
      new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster %in% setdiff(old_centroids$WCCluster, new_centroids$WCCluster),])
    }
  }

  return(new_centroids)
}

#' Solution for Cluster Representative update which uses Midrange.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_midrange <- function(data, assignment, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  new_centroids <- stats::aggregate(x = data, by = list(assignment), FUN = function(x) {(stats::quantile(x)[[1]] + stats::quantile(x)[[5]])/2})
  colnames(new_centroids)[1] <- "WCCluster"

  #If some cluster does not have assignments, add old one
  if(!is.null(old_centroids))
  {
    if(nrow(new_centroids) != nrow(old_centroids))
    {
      new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster %in% setdiff(old_centroids$WCCluster, new_centroids$WCCluster),])
    }
  }

  return(new_centroids)
}

#' Solution for Cluster Representative update which uses Online mean.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param assignment_type Assignment type to be used.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_online_mean <- function(data, assignment, assignment_type, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  assignments <- assignment
  for(i in 1:nrow(data))
  {
    if(i == 1)
    {
      new_centroids <- wc_recalc_mean(data = data, assignment = assignments, old_centroids = old_centroids)
    }
    new_assignment <- wc_assignment(data = data, centroids = new_centroids, assignment_type = assignment_type)
    new_centroids <- wc_recalc_mean(data = data, assignment = new_assignment, old_centroids = old_centroids)

    #IF ASSIGNMENT WAS EMPTY
    k <- nrow(new_centroids)
    for(cent in 1:k)
    {
      if(nrow(new_centroids[new_centroids$WCCluster == cent, ]) == 0)
      {
        new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster == cent, ])
      }
      new_centroids <- new_centroids[order(new_centroids$WCCluster, decreasing = FALSE), ]
    }

    old_centroids <- new_centroids

    if(sum(new_assignment == assignments) == length(assignments))
    {
      break
    }
    else
    {
      assignments <- new_assignment
    }
  }
  return(new_centroids)
}

#' Solution for Cluster Representative update which uses Online median.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param assignment_type Assignment type to be used.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_online_median <- function(data, assignment, assignment_type, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  assignments <- assignment
  for(i in 1:nrow(data))
  {
    if(i == 1)
    {
      new_centroids <- wc_recalc_median(data = data, assignment = assignments, old_centroids = old_centroids)
    }
    new_assignment <- wc_assignment(data = data, centroids = new_centroids, assignment_type = assignment_type)
    new_centroids <- wc_recalc_median(data = data, assignment = new_assignment, old_centroids = old_centroids)

    #IF ASSIGNMENT WAS EMPTY
    k <- nrow(new_centroids)
    for(cent in 1:k)
    {
      if(nrow(new_centroids[new_centroids$WCCluster == cent, ]) == 0)
      {
        new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster == cent, ])
      }
      new_centroids <- new_centroids[order(new_centroids$WCCluster, decreasing = FALSE), ]
    }

    old_centroids <- new_centroids

    if(sum(new_assignment == assignments) == length(assignments))
    {
      break
    }
    else
    {
      assignments <- new_assignment
    }
  }
  return(new_centroids)
}

#' Solution for Cluster Representative update which uses Online Trimmed mean.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param assignment_type Assignment type to be used.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_online_trimmed_mean <- function(data, assignment, assignment_type, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  assignments <- assignment
  for(i in 1:nrow(data))
  {
    if(i == 1)
    {
      new_centroids <- wc_recalc_trimmed_mean(data = data, assignment = assignments, old_centroids = old_centroids)
    }
    new_assignment <- wc_assignment(data = data, centroids = new_centroids, assignment_type = assignment_type)
    new_centroids <- wc_recalc_trimmed_mean(data = data, assignment = new_assignment, old_centroids = old_centroids)

    #IF ASSIGNMENT WAS EMPTY
    k <- nrow(new_centroids)
    for(cent in 1:k)
    {
      if(nrow(new_centroids[new_centroids$WCCluster == cent, ]) == 0)
      {
        new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster == cent, ])
      }
      new_centroids <- new_centroids[order(new_centroids$WCCluster, decreasing = FALSE), ]
    }

    old_centroids <- new_centroids

    if(sum(new_assignment == assignments) == length(assignments))
    {
      break
    }
    else
    {
      assignments <- new_assignment
    }
  }
  return(new_centroids)
}

#' Solution for Cluster Representative update which uses Online Geometric mean.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param assignment_type Assignment type to be used.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_online_geometric_mean <- function(data, assignment, assignment_type, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  assignments <- assignment
  for(i in 1:nrow(data))
  {
    if(i == 1)
    {
      new_centroids <- wc_recalc_geometric_mean(data = data, assignment = assignments, old_centroids = old_centroids)
    }
    new_assignment <- wc_assignment(data = data, centroids = new_centroids, assignment_type = assignment_type)
    new_centroids <- wc_recalc_geometric_mean(data = data, assignment = new_assignment, old_centroids = old_centroids)

    #IF ASSIGNMENT WAS EMPTY
    k <- nrow(new_centroids)
    for(cent in 1:k)
    {
      if(nrow(new_centroids[new_centroids$WCCluster == cent, ]) == 0)
      {
        new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster == cent, ])
      }
      new_centroids <- new_centroids[order(new_centroids$WCCluster, decreasing = FALSE), ]
    }

    old_centroids <- new_centroids

    if(sum(new_assignment == assignments) == length(assignments))
    {
      break
    }
    else
    {
      assignments <- new_assignment
    }
  }
  return(new_centroids)
}

#' Solution for Cluster Representative update which uses Online Harmonic mean.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param assignment_type Assignment type to be used.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_online_harmonic_mean <- function(data, assignment, assignment_type, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  assignments <- assignment
  for(i in 1:nrow(data))
  {
    if(i == 1)
    {
      new_centroids <- wc_recalc_harmonic_mean(data = data, assignment = assignments, old_centroids = old_centroids)
    }
    new_assignment <- wc_assignment(data = data, centroids = new_centroids, assignment_type = assignment_type)
    new_centroids <- wc_recalc_harmonic_mean(data = data, assignment = new_assignment, old_centroids = old_centroids)

    #IF ASSIGNMENT WAS EMPTY
    k <- nrow(new_centroids)
    for(cent in 1:k)
    {
      if(nrow(new_centroids[new_centroids$WCCluster == cent, ]) == 0)
      {
        new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster == cent, ])
      }
      new_centroids <- new_centroids[order(new_centroids$WCCluster, decreasing = FALSE), ]
    }

    old_centroids <- new_centroids

    if(sum(new_assignment == assignments) == length(assignments))
    {
      break
    }
    else
    {
      assignments <- new_assignment
    }
  }
  return(new_centroids)
}

#' Solution for Cluster Representative update which uses Online Quadratic mean.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param assignment_type Assignment type to be used.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_online_quadratic_mean <- function(data, assignment, assignment_type, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  assignments <- assignment
  for(i in 1:nrow(data))
  {
    if(i == 1)
    {
      new_centroids <- wc_recalc_quadratic_mean(data = data, assignment = assignments, old_centroids = old_centroids)
    }
    new_assignment <- wc_assignment(data = data, centroids = new_centroids, assignment_type = assignment_type)
    new_centroids <- wc_recalc_quadratic_mean(data = data, assignment = new_assignment, old_centroids = old_centroids)

    #IF ASSIGNMENT WAS EMPTY
    k <- nrow(new_centroids)
    for(cent in 1:k)
    {
      if(nrow(new_centroids[new_centroids$WCCluster == cent, ]) == 0)
      {
        new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster == cent, ])
      }
      new_centroids <- new_centroids[order(new_centroids$WCCluster, decreasing = FALSE), ]
    }

    old_centroids <- new_centroids

    if(sum(new_assignment == assignments) == length(assignments))
    {
      break
    }
    else
    {
      assignments <- new_assignment
    }
  }
  return(new_centroids)
}

#' Solution for Cluster Representative update which uses Online Trimean.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param assignment_type Assignment type to be used.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_online_trimean <- function(data, assignment, assignment_type, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  assignments <- assignment
  for(i in 1:nrow(data))
  {
    if(i == 1)
    {
      new_centroids <- wc_recalc_trimean(data = data, assignment = assignments, old_centroids = old_centroids)
    }
    new_assignment <- wc_assignment(data = data, centroids = new_centroids, assignment_type = assignment_type)
    new_centroids <- wc_recalc_trimean(data = data, assignment = new_assignment, old_centroids = old_centroids)

    #IF ASSIGNMENT WAS EMPTY
    k <- nrow(new_centroids)
    for(cent in 1:k)
    {
      if(nrow(new_centroids[new_centroids$WCCluster == cent, ]) == 0)
      {
        new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster == cent, ])
      }
      new_centroids <- new_centroids[order(new_centroids$WCCluster, decreasing = FALSE), ]
    }

    old_centroids <- new_centroids

    if(sum(new_assignment == assignments) == length(assignments))
    {
      break
    }
    else
    {
      assignments <- new_assignment
    }
  }
  return(new_centroids)
}

#' Solution for Cluster Representative update which uses Online Midhindge.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param assignment_type Assignment type to be used.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_online_midhinge <- function(data, assignment, assignment_type, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  assignments <- assignment
  for(i in 1:nrow(data))
  {
    if(i == 1)
    {
      new_centroids <- wc_recalc_midhinge(data = data, assignment = assignments, old_centroids = old_centroids)
    }
    new_assignment <- wc_assignment(data = data, centroids = new_centroids, assignment_type = assignment_type)
    new_centroids <- wc_recalc_midhinge(data = data, assignment = new_assignment, old_centroids = old_centroids)

    #IF ASSIGNMENT WAS EMPTY
    k <- nrow(new_centroids)
    for(cent in 1:k)
    {
      if(nrow(new_centroids[new_centroids$WCCluster == cent, ]) == 0)
      {
        new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster == cent, ])
      }
      new_centroids <- new_centroids[order(new_centroids$WCCluster, decreasing = FALSE), ]
    }

    old_centroids <- new_centroids

    if(sum(new_assignment == assignments) == length(assignments))
    {
      break
    }
    else
    {
      assignments <- new_assignment
    }
  }
  return(new_centroids)
}

#' Solution for Cluster Representative update which uses Online Midrange.
#'
#' @param data A dataset for which Cluster Representatives needs to be updated.
#' @param assignment Vector of Cluster assignments.
#' @param assignment_type Assignment type to be used.
#' @param old_centroids Old centroids.
#' @return As a result new Cluster Representatives are obtained. Result is in for of data.frame or data.matrix.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalc_online_midrange <- function(data, assignment, assignment_type, old_centroids = NULL)
{
  if(!(class(data) %in% c('data.frame', 'matrix')))
  {
    stop('Data should be data.frame or matrix')
  }

  if(class(assignment) != 'numeric')
  {
    stop('Assignment should be numeric vector')
  }

  if(nrow(data) != length(assignment))
  {
    stop('Data and Assignments are not in compliance')
  }

  assignments <- assignment
  for(i in 1:nrow(data))
  {
    if(i == 1)
    {
      new_centroids <- wc_recalc_midrange(data = data, assignment = assignments, old_centroids = old_centroids)
    }
    new_assignment <- wc_assignment(data = data, centroids = new_centroids, assignment_type = assignment_type)
    new_centroids <- wc_recalc_midrange(data = data, assignment = new_assignment, old_centroids = old_centroids)

    #IF ASSIGNMENT WAS EMPTY
    k <- nrow(new_centroids)
    for(cent in 1:k)
    {
      if(nrow(new_centroids[new_centroids$WCCluster == cent, ]) == 0)
      {
        new_centroids <- rbind.data.frame(new_centroids, old_centroids[old_centroids$WCCluster == cent, ])
      }
      new_centroids <- new_centroids[order(new_centroids$WCCluster, decreasing = FALSE), ]
    }

    old_centroids <- new_centroids

    if(sum(new_assignment == assignments) == length(assignments))
    {
      break
    }
    else
    {
      assignments <- new_assignment
    }
  }
  return(new_centroids)
}

#' Data frame for possible values of recalculate types.
#' @author Sandro Radovanovic \email{sandro.radovanovic@@gmail.com}
wc_recalculate_types <- data.frame()

wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Mean', 'Method' = 'wc_recalc_mean'))
wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Median', 'Method' = 'wc_recalc_median'))
wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Trimmed mean', 'Method' = 'wc_recalc_trimmed_mean'))
wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Geometric mean', 'Method' = 'wc_recalc_geometric_mean'))
wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Harmonic mean', 'Method' = 'wc_recalc_harmonic_mean'))
wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Quadratic mean', 'Method' = 'wc_recalc_quadratic_mean'))
wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Trimean', 'Method' = 'wc_recalc_trimean'))
wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Midhinge', 'Method' = 'wc_recalc_midhinge'))
wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Midrange', 'Method' = 'wc_recalc_midrange'))
wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Online mean', 'Method' = 'wc_recalc_online_mean'))
wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Online median', 'Method' = 'wc_recalc_online_median'))
wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Online trimmed mean', 'Method' = 'wc_recalc_online_trimmed_mean'))
wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Online geometric mean', 'Method' = 'wc_recalc_online_geometric_mean'))
wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Online harmonic mean', 'Method' = 'wc_recalc_online_harmonic_mean'))
wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Online Quadratic mean', 'Method' = 'wc_recalc_online_quadratic_mean'))
wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Online trimean', 'Method' = 'wc_recalc_online_trimean'))
wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Online midhinge', 'Method' = 'wc_recalc_online_midhinge'))
wc_recalculate_types <- rbind.data.frame(wc_recalculate_types, data.frame('Type' = 'Online midrange', 'Method' = 'wc_recalc_online_midrange'))