This Java package performs cluster center initialization (CCIA) for K-means algorithms. The implementation is based on the following paper -

Cluster Center Initialization Algorithm for K-mean Clustering, Shehroz S. Khan and Amir Ahmad, Pattern Recognition Letters, Volume 25, No. 11, pages 1293-1302, 2004

The paper is available at - https://pdfs.semanticscholar.org/0288/181f90c5f85ba219ebc4beb7c759fd052408.pdf

Author: Shehroz S. Khan

Affiliation: University of Toronto, Canada

About: The program generates fixed cluster centers that are written to a separate file. These centers are then used by K-means clustering and the data object membership is written in another file.

This distribution contains two java files:

1. KmeansClustering.java -- This is the main class file to generate initial cluster centers.

2. initKmeans.java -- This is the test file that uses KmeansClustering.java to generate initial modes and perform K-means clustering afterwards.

There are three things that needs to be set before executing the initKmeans.java

• The path of source file name. The file name should be in arff format.

  String targetDir = "//path//to//directory//for//data//"; //Directory Name

  String inputFile = targetDir+"iris.arff"; // Data file name

  The output files for initial centers and data object memberships are written with subscript "-centers.csv" and "-membership.csv" to the input file name.

• The number of clusters in the data. It can be done by altering this line

  initkm.setK(N);//number of clusters

  where N is the number of clusters in the data

• The number of nearest neighbours to merge cluster center strings to arrive at K initial clusters.

  km1.setNN(1); //for merging clusters, distance of 'q^th' nearest neighbour of a cluster center.

  //For practical purposes a value of 1 is good. Increasing this value to large number will result in exception.

If eclipse is not used then the following line can be removed from the top

package initKmean;

Dependencies

This package is tested using Weka-3-9-2, Apache Commons Math 3.6.1 and Apache Common Lang 3.3.7. The respective jars can be obtained from the following hyperlinks:

https://www.cs.waikato.ac.nz/ml/weka/downloading.html

http://commons.apache.org/proper/commons-math/download_math.cgi

http://commons.apache.org/proper/commons-lang/download_lang.cgi

Citation

If you use this program in your research and publish a paper, then please use the following citation

@article{khan2004cluster, title={Cluster center initialization algorithm for K-means clustering}, author={Khan, Shehroz S and Ahmad, Amir}, journal={Pattern recognition letters}, volume={25}, number={11}, pages={1293--1302}, year={2004},
publisher={Elsevier} }

ccia