Rcmd
Clustering
Summary
Rcmd::Clustering - Interfaces to clustering algorithms of R language.
Package variables
No package variables defined.
Included modules
Synopsis
No synopsis!
Description
This class is a part of G-language Genome Analysis Environment,
collecting interfaces to clustering algorithms of R language.
Methods
| hclust | No description | Code |
| kmeans | No description | Code |
| sample_data_for_clustering | No description | Code |
| set_clust_data | No description | Code |
| som | No description | Code |
Methods description
None available.
Methods code
sub hclust
{ my $rcmd = shift;
$rcmd->set_mode('tmp');
opt_default("dist_method"=>"euclidean", "hclust_method"=>"com",
filename=>"hclust.png", output=>"show", sampledata=>0);
my @args = opt_get(@_);
my $dist_method = opt_val("dist_method");
my $hclust_method = opt_val("hclust_method");
my $filename= opt_val("filename");
my $output = opt_val("output");
my $label = opt_val("label");
my $sampledata = opt_val("sampledata");
if($sampledata){
$rcmd->sample_data_for_clustering();
}else{
$rcmd->set_clust_data(@args, -label=>$label);
}
$rcmd->exec(
'require(mva)',
"rclust.hclust<-hclust(dist(rclust,\" $dist_method\"),\" $hclust_method\")"
);
my @result = $rcmd->exec('rclust.hclust$merge');
if($output =~ /g/ || $output =~ /show/){
$rcmd->exec(
'postscript("/tmp/rclust.ps")',
'plot(rclust.hclust,hang=-1)'
);
mkdir('graph', 0777);
system("convert -rotate 90 /tmp/rclust.ps graph/$filename");
msg_gimv("graph/$filename") if ($output =~ /show/);
}
shift @result;
$rcmd->set_mode();
return @result;} | sub kmeans
{ my $rcmd = shift;
$rcmd->set_mode('tmp');
opt_default("iter.max"=>10, filename=>"kmeans.png", output=>"show", sampledata=>0, centers=>5);
my @args = opt_get(@_);
my $centers = opt_val("centers");
my $iter = opt_val("iter.max");
my $output = opt_val("output");
my $filename = opt_val("filename");
my $sampledata = opt_val("sampledata");
my $label = opt_val("label");
if($sampledata){
$rcmd->sample_data_for_clustering();
}else{
$rcmd->set_clust_data(@args, -label=>$label);
}
$rcmd->exec(
'require(mva)',
"rclust.kmeans<-kmeans(rclust,$centers,$iter)"
);
my @result = $rcmd->exec('rclust.kmeans$cluster');
if($output =~ /g/ || $output =~ /show/){
$rcmd->exec(
'postscript("/tmp/kmeans.ps")',
'plot(rclust,col=rclust.kmeans$cluster)',
"points(rclust.kmeans\$centers, col=1:$centers,pch=8)"
);
mkdir('graph', 0777);
system("convert -rotate 90 /tmp/kmeans.ps graph/$filename");
msg_gimv("graph/$filename") if ($output =~ /show/);
}
$rcmd->set_mode();
return @result;} | sub sample_data_for_clustering
{ my $rcmd = shift;
$rcmd->exec(
'rclust<-rbind(matrix(rnorm(100,sd=0.3),ncol=2),' .
'matrix(rnorm(100,mean=1,sd=0.3),ncol=2))'
);} | sub set_clust_data
{ my $rcmd = shift;
my @args = opt_get(@_);
my $label = opt_val("label");
my $flag = 0;
foreach (@args){
if ($flag == 0){
$flag ++;
$rcmd->array('rclust', @$_);
}else{
$rcmd->array('tmp', @$_);
$rcmd->exec('rclust <- cbind(rclust, tmp)');
}
}
if (length $label){
$rcmd->sarray('label', @$label);
$rcmd->exec('dimnames(rclust) <- list(label, NULL)');
}} | sub som
{ my $rcmd = shift;
$rcmd->set_mode('tmp');
opt_default(filename=>"som.png", output=>"show", xdim=>3, ydim=>3, sampledata=>0);
my @args = opt_get(@_);
my $xdim = opt_val("xdim");
my $ydim = opt_val("ydim");
my $filename= opt_val("filename");
my $output = opt_val("output");
my $sampledata = opt_val("sampledata");
my $label = opt_val("label");
if($sampledata){
$rcmd->sample_data_for_clustering();
}else{
$rcmd->set_clust_data(@args, -label=>$label);
}
my $message = $rcmd->exec(
'require(GeneSOM)'
);
croak(
"\n\nPackage GeneSOM not installed.\nrun\n\tinstall.packages(\'GeneSOM\')\n" .
"in R as root.\n\n"
) if ($message =~ /FALSE/);
$rcmd->exec(
"rclust.som<-som(rclust, $xdim, $ydim)",
'rclust.som$visual'
);
my @result;
open(FILE, $rcmd->{log}) || die($!);
while(<FILE>){
if(/qerror/){
@result = ();
while(<FILE>){
chomp;
if(/^\d/){
my (undef, $x, $y, $qerror) = split(/\s+/, $_, 4);
push(@result, [$x, $y, $qerror]);
}
}
}
}
close(FILE);
if($output =~ /g/ || $output =~ /show/){
$rcmd->exec(
'postscript("/tmp/som.ps")',
'plot(rclust.som)'
);
mkdir('graph', 0777);
system("convert -rotate 90 /tmp/som.ps graph/$filename");
msg_gimv("graph/$filename") if ($output =~ /show/);
}
$rcmd->set_mode();
return @result;} | General documentation
Name: $rcmd->som() - clustering using Self-Organizing Map
Description:
Clustering with Self-Organizing Map (SOM) using R language.
Installation of GeneSOM library for R language is required.
run R as a super user - sudo R - and type the following:
install.packages('GeneSOM'))
Returns a two-dimensional array correspondingn to the
result$visual of som() in R's GeneSOM library.
Usage:
$rcmd = new Rcmd();
@result = $rcmd->som(\@array1, \@array2, \@array3, ..., -label=>\@label);
or
@result = $rcmd->som(-sampledata=>1);
Arrays correspond to the columns (data series), and labels for each of
these arrays can be given by -label option.
Options:
-label labels or names of the data series.
-xdim x-dimension of the map (default: 3)
-ydim y-dimension of the map (default: 3)
-filename output filename of the graph (default: som.png)
-output output toggle option (default: show)
"g" to generate graph without displaying.
-sampledata use sample data (default: 0)
Author:
Kazuharu Arakawa (gaou@sfc.keio.ac.jp)
History:
20070612-01 converted to Rcmd::Clustering
20030904-01 initial posting (G::Tools::RCluster) Name: $rcmd->hclust() - hierarchical clustering analysis
Description:
Hierarchical clustering using R language.
Distance measure to use can be specified by -dist_method option.
(euclidean:default, maximum, manhattan, canberra, binary)
Clustering method can be specified by -hclust_method option.
(complete:default, ward, single, average, mcquitty, median, centroid)
This method automatically runs dist() in R to calculate the dissimilarities
of the given data and passes the result to hclust() of R.
Returned value corresponds to result$merge of hclust() in R (n-1 by 2 matrix).
Row i describes the merging of clusters at step i.
Usage:
$rcmd = new Rcmd();
@merge = $rcmd->hclust(\@array1, \@array2, \@array3, ..., -label=>\@label);
or
@merge = $rcmd->hclust(-sampledata=>1);
Arrays correspond to the columns (data series), and labels for each of
these arrays can be given by -label option.
Options:
-label labels or names of the data series.
-dist_method distance measure (euclidean:default, maximum, manhattan,
canberra, binary)
-hclust_method clustering method (complete:default, ward, single, average,
mcquitty, median, centroid)
-filename output filename of the graph (default: hclust.png)
-output output toggle option (default: show)
"g" to generate graph without displaying.
-sampledata use sample data (default: 0)
Author:
Kazuharu Arakawa (gaou@sfc.keio.ac.jp)
History:
20070612-01 converted to Rcmd::Clustering
20030904-01 initial posting (G::Tools::RCluster) Name: $rcmd->kmeans() - clustering with K-means method
Description:
Clustering with K-means method with using R language.
Number of cluster centers can be given by -centers option (default: 10)
and number of iterations is given by -iter.max (default: 10).
Returned value corresponds to result$cluster of kmeans() in R.
(a vector of cluster numbers to which each point is allocated)
Usage:
$rcmd = new Rcmd();
@cluster = $rcmd->kmeans(\@array1, \@array2, \@array3, ..., -label=>\@label);
or
@cluster = $rcmd->kmeans(-sampledata=>1);
Arrays correspond to the columns (data series), and labels for each of
these arrays can be given by -label option.
Options:
-label labels or names of the data series.
-center number of cluster centers (default: 5)
-iter.max number of iterations (default: 10)
-filename output filename of the graph (default: hclust.png)
-output output toggle option (default: show)
"g" to generate graph without displaying.
-sampledata use sample data (default: 0)
Author:
Kazuharu Arakawa (gaou@sfc.keio.ac.jp)
History:
20070612-01 converted to Rcmd::Clustering
20030904-01 initial posting (G::Tools::RCluster)