Home // INTELLI 2014, The Third International Conference on Intelligent Systems and Applications // View article

Refining the Scatteredness of Classes using Pheromone-Based Kohonen Self-Organizing Map (PKSOM)

Authors:
Azlin Ahmad
Rubiyah Yusof

Keywords: Kohonen Self-Organizing Map (KSOM); Pheromone; Ant Clustering Algorithm (ACA); Clustering; Cluster density.

Abstract:
The Kohonen Self-Organizing Map (KSOM) is one of the well-known unsupervised learning algorithms, which has been applied in various areas. This algorithm can cluster and classify an enormous amount of data into several clusters according to the similarity of the data features. However, it has many drawbacks, such as difficulty of clustering the data, which have similar features. These may lead to the inefficient result; the data is scatteredly mapped even though it is accurately clustered into several clusters according to the features. Therefore, this paper proposed a Pheromone-based Kohonen Self-Organizing Map (PKSOM) algorithm to refine the scatteredness of the data in the clusters, thus to improve the cluster density. Some modifications have been made to the original Kohonen Self-Organizing Map (KSOM), adapted from the Ant Clustering Algorithm procedures. This PKSOM has been tested on three different datasets; Iris flowers, Glass and Wood datasets. Based on the result, the proposed method has improved the classification impressively by increasing the density of the data in clusters. Hence, it has also refined the scatteredness of classes, where each dataset is well clustered where data that have similar features are located closely to each other in the same cluster. However, there are a few overlapped clusters that still occur.

Pages: 107 to 113

Copyright: Copyright (c) IARIA, 2014

Publication date: June 22, 2014

Published in: conference

ISSN: 2308-4065

ISBN: 978-1-61208-352-0

Location: Seville, Spain

Dates: from June 22, 2014 to June 26, 2014