cc_coclus: CoClus

In Rcoclust: Co-Clustering with Document-Term Matrix

Description

Implementation of the method which aims at a graph modularity maximization. The data vectors for the columns are not required for the input parameters in envrdata.

Usage

cc_coclus(g,envrdata,zi,wj,transfrm,maxiter,debug)

Arguments

g	number of clusters for rows and columns.
envrdata	environment with data.
zi	row clusters.
wj	column clusters.
transfrm	transformation of the data (0:none, 1:binarization, 2:tf-idf, 3:tf-idf+rows normalization).
maxiter	maximum number of iterations.
debug	flag for debug, if equal to 1 shows some informations to user.

Value

The function alters the parameters zi and wj. It returns a vector with the value of the objective function per iterations, and zi, wj which are the estimated cluster labels.

Examples

  library(Rcoclust);
  
  #load data
  data(data_news4);
  envrdata=get_envrdata(A_ijx,lbs,name,0);
  
  #retrieve matrix size and number of classes
  n=envrdata$n;
  d=envrdata$d;
  g=length(unique(envrdata$lbs));
  
  #coclus
  bestresu=NULL;
  for (m in 1:100) {
    zi_coclus=as.integer( sample(x = 1:g,size = n,replace = TRUE) );
    wj_coclus=as.integer( sample(x = 1:g,size = d,replace = TRUE) );
    resu=Rcoclust::cc_coclus(g,envrdata,zi_coclus,wj_coclus,3,80,0);
    if (m==1) bestresu=resu;
    if (m>1) {
      if (resu$obj[length(resu$obj)]>bestresu$obj[length(bestresu$obj)])
        bestresu=resu;
    }
  }
  print(table(envrdata$lbs,bestresu$zi));

Rcoclust documentation built on May 6, 2017, 3:34 a.m.