R/ClusterRedefine.R

In FCPS: Fundamental Clustering Problems Suite

ClusterRedefine=function(Cls,NewLabels,OldLabels,Silent=FALSE){
  # 
  # INPUT
  # Cls                1:n numerical vector of numbers defining the classification as the main
  #                    output of the clustering algorithm for the n cases of data. It has k unique
  #                    numbers representing the arbitrary labels of the clustering.
  # NewLabels        [1:p], p<=k identifiers of clusters to be changed with
  #Optional
  # OldLabels    [1:p], p<=k identifiers of clusters to be changed, default [1:k] unique cluster Ids of cls
  # 
  # OUTPUT
  # Cls[1:n] numerical vector named after NewLabels 
  # 
    if(!is.vector(Cls)){
      if(isFALSE(Silent))
        warning('ClusterRedefine: Cls is not a vector. Calling as.numeric(as.character(Cls))')
      Cls=as.numeric(as.character(Cls))
    }
  if(missing(OldLabels))
    OldLabels=unique(Cls)
  
  if(length(OldLabels)!=length(NewLabels)){
    if(isFALSE(Silent))
      warning('ClusterRedefine: length(OldLabels)!=length(NewLabels))')
    
    return(Cls)
  }
  if(length(unique(Cls))==1){
    if(isFALSE(Silent))
      warning("ClusterRedefine: Only one unique label in Cls given.")
    #return(Cls)
  }
  # #store old clusters
  # indV=list()
  # for(i in 1:length(OldLabels)){#iterate cluster labels
  #   #stores boolean vector
  #   indV[[i]]=c(Cls==OldLabels[i])
  # }
  # #apply renaming but ignore same label in old cluster and new cluster that indicates different clusters!
  # for(i in 1:length(OldLabels)){
  #   Cls[indV[[i]]]=NewLabels[i]
  # }
  #faster version
  map <- setNames(NewLabels, OldLabels)      # a named character vector
  cls_chr <- as.character(Cls)
  mapped <- map[cls_chr]     # NA where no match
  #should not happencause of error catch above
  #this takes too long for large vectors
  #Cls <- ifelse(is.na(mapped), cls_chr, mapped)
  Cls=unname(mapped)
  return(Cls)
}