R/trajReduce.R

In traj: Feature-Based Clustering of Longitudinal Trajectories

#'@title Select a Subset of the Measures Using a Similarity Index on the Set of Clusterings
#'
#'@description This function examines the effect of reducing the number of measures on which the trajectories are clustered. Specifically, starting from a clustering \eqn{C} in the form of an object of class \code{trajClusters} and a choice of a similarity index to compare clusterings, this function finds the subset of measures which results in the clustering most similar to \eqn{C}.
#'
#'@param Measures object of class \code{trajMeasures} as returned by \code{\link[traj]{trajMeasures}}.
#'@param Clusters object of class \code{trajClusters} as returned by \code{\link[traj]{trajClusters}}.
#'@param index The similarity index. Either "ARI" for the Adjusted Rand Index of Hubert and Arabie (1985), "nVId" for the normalized variation of information distance (eg. Meila (2007)) or "nSJd" for the normalized split/joint distance of van Dongen (2000).
#'@param keep The number of measures to keep. Defaults to 3.
#'
#'@details The Rand index ranges from 0 to 1 with 0 indicating identical clusters and 1 indicating maximally different clusters. The normalized variation of information distance (nVId) and normalized split-join distance (nSJd) and have the opposite interpretation with 0 indicating maximally different clusters and 1 indicating identical clusters. Therefor, to facilitate comparison, we plot 1 - nVId (resp. 1 - nSJd) instead of nVId (resp. nSJd).
#'
#'@importFrom igraph compare
#'
#'@references 
#'Hubert L, Arabie P. Comparing partitions. Journal of Classification 2:193-218, 1985.
#'
#'Meila M. Comparing clusterings -- an information based distance. Journal of Multivariate Analysis, 98, pp 873-895, 2007.
#'  
#'van Dongen S. Performance criteria for graph clustering and Markov cluster experiments. Technical Report INS-R0012, National Research Institute for Mathematics and Computer Science in the Netherlands, Amsterdam, May 2000.
#'  
#'
#' @examples
#' \dontrun{
#'data("trajdata")
#'trajdata.noGrp <- trajdata[, -which(colnames(trajdata) == "Group")] #remove the Group column
#'
#'m = trajMeasures(trajdata.noGrp, ID = TRUE)
#'trajReduce(m)
#'}
#'
#'@rdname trajReduce
#'
#'@export
trajReduce <-
  function (Measures, Clusters, index = "ARI", keep = 3) {
    
    if(is.null(Clusters$nclusters)){
      stop("The 'Clusters' argument must be the output of the 'trajClusters' function in which the 'nclusters' argument is not 'NULL'.")
    } else {
      
      useful.fct <- function(list_of_vectors){
        output.list <- list()
        N <- length(list_of_vectors)
        for(i in 1:N){
          ith.vector <- list_of_vectors[[i]]
          n.i <- length(ith.vector)
          for(j in 1:n.i){
            output.list[[length(output.list)+1]] <- ith.vector[-j]
          }
        }
        return(output.list)
      }
      
      traj <- as.factor(Clusters$partition[,"Cluster"])
      nCluster <- Clusters$nclusters
      n <- nrow(Clusters$partition)
      
      list.of.tables <- list()
      
      par(mar=c(5.1, 5.1, 4.1, 2.1), lwd = 3, cex.axis = 2, cex.lab = 2, cex.main = 2, bty = "o", fg = "black") 
      par(mfrow = c(1, length(index)))
      
      combin <- utils::combn(Clusters$select, keep)
      
      criterion.v <- c()
      
      for(i in 1:(ncol(combin))){
        s3.i = quiet(trajClusters(Measures, select = combin[, i], fuzzy = FALSE, nclusters =  Clusters$nclusters))
        
        traj.i <- as.factor(s3.i$partition[,"Cluster"])
        
        if (index == "ARI") {
          criterion.v <- c(criterion.v, igraph::compare(traj, traj.i, method = "adjusted.rand"))
        }
        if (index == "nVId") {
          criterion.v <- c(criterion.v, 1-igraph::compare(traj, traj.i, method = "vi")/(2*log(nCluster)))
        }
        if (index == "nSJd") {
          criterion.v <- c(criterion.v, 1-igraph::compare(traj, traj.i, method = "split.join") / (2*nCluster*(n/nCluster - ceiling(n/(nCluster^2)))) )
        }
      }
      
      w <- which(criterion.v == max(criterion.v))[1]
      
      Clusters.red <- quiet(trajClusters(Measures, select = combin[, w], fuzzy = Clusters$fuzzy, nclusters = keep))
      
      traj.red <- as.factor(Clusters.red$partition[, "Cluster"])
      
      output <- list(
        reduced.list = combin[, w],
        index.value = criterion.v[w][1],
        partition.red = traj.red,
        fuzzy.partition.red = Clusters.red$fuzzy.partition
      )
      
      return(output)
    }
  }