NAME
Statistics::Gap - Perl extension for the "Gap Statistics"
SYNOPSIS
use Statistics::Gap;
&gap("GapPrefix", "Filename.txt", "manhattan", "agglo", 5, 3);
DESCRIPTION
Given a dataset how does one automatically find the optimal number
of clusters that the dataset should be grouped into? - is one of the
prevailing problems. Statisticians Robert Tibshirani, Guenther Walther
and Trevor Hastie propose a solution for this problem is a Techinal
Report named - "Estimating the number of clusters in a dataset via
the Gap Statistics". This perl module implements the approach proposed
in the above paper.
EXPORT
"gap" function by default.
INPUT
Prefix
The string that should be used to as a prefix while naming the
intermediate files and the .png files (graph files).
InputFile
The input dataset is expected in a plain text file where the first
line in the file gives the dimensions of the dataset and then the
dataset in a matrix format should follow. The contexts / observations
should be along the rows and the features should be along the column.
DistanceMeasure
The Distance Measure that should be used.
Currrently this module supports the following distance measure:
1. Manhattan (string that should be used as an argument: "manhattan")
2. Euclidean (string that should be used as an argument: "euclidean")
3. Squared Euclidean (string that should be used as an argument: "squared")
ClusteringAlgorithm
The Clustering Measures that can be used are:
1. rb - Repeated Bisections [Default]
2. rbr - Repeated Bisections for by k-way refinement
3. direct - Direct k-way clustering
4. agglo - Agglomerative clustering
5. graph - Graph partitioning-based clustering
6. bagglo - Partitional biased Agglomerative clustering
K value
This is an approximate upper bound for the number of clusters that may be
present in the dataset. Thus for a dataset that you expect to be seperated
into 3 clusters this value should be set some integer value greater than 3.
B value
Specifies the number of time the reference distribution should be generated
Typical value would be 3.
OUTPUT
The output returned is a single integer value which indicates the optimal
number of clusters that the input dataset should be clustered into.
PRE-REQUISITES
This module uses suite of C programs called CLUTO for clustering purposes.
Thus CLUTO needs to be installed for this module to be functional.
CLUTO can be downloaded from http://www-users.cs.umn.edu/~karypis/cluto/
SEE ALSO
http://citeseer.ist.psu.edu/tibshirani00estimating.html
http://www-users.cs.umn.edu/~karypis/cluto/
AUTHOR
Anagha Kulkarni, University of Minnesota Duluth
kulka020 d.umn.edu
Ted Pedersen, University of Minnesota Duluth
tpederse d.umn.edu
Guergana Savova, Mayo Clinic
savova.guergana mayo.edu
COPYRIGHT AND LICENSE
Copyright (C) 2005-2006, Ted Pedersen, Guergana Savova and Anagha Kulkarni
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; either version 2
of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.