/ 
Statistics-Gap-0.01
River stage zero No dependents
/ Statistics::Gap

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 number 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 <at> d.umn.edu
	
    Ted Pedersen, University of Minnesota Duluth
    tpederse <at> d.umn.edu

    Guergana Savova, Mayo Clinic
    savova.guergana <at> 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.