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 Guergana Savova, Mayo Clinic savova.guergana mayo.edu COPYRIGHT AND LICENSE Copyright (C) 2005-2006, 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.