Adaptive clustering with artificial ants

Abstract

Clustering task aims at the unsupervised classification of patterns (e.g., observations, data, vec- tors, etc.) in different groups. Clustering problem has been approached from different disciplines during the last years. Although have been proposed different alternatives to cope with clustering, there also exists an interesting and novel field of research from which different bioinspired algorithms have emerged, e.g., genetic algorithms and ant colony algorithms. In this article we pro- pose an extension of the AntTree algorithm, an example of an algorithm recently proposed for a data mining task which is designed following the principle of self-assembling behavior observed in some species of real ants. The extension proposed called Adaptive-AntTree (AAT for short) represents a more flexible version of the original one. The ants in AAT are able of changing the assigned position in previous iterations in the tree under construction. As a consequence, this new algorithm builds an adaptive hierarchical cluster which changes over the run in order to improve the final result. The AAT performance is experimentally analyzed and compared against AntTree and K-means which is one of the more popular and referenced clustering algorithm.