SwarmClass: A Novel Data Clustering Approach by a Hybridization of an Ant Colony with Flying Insects

Abstract

Swarm behaviors contribute to the resolution of very large number of difficult tasks thanks to simplified models and elementary rules [1]. This work claims a new swarm based behavior used for unsupervised classification. The proposed behavior starts from the ants collective sorting behavior as initially proposed by Lumer and Faieta [2] and overwrites it with additional behaviors inspired from birds and spiders. Our algorithm is then based on the existing work of [3], [4] and [2]. The proposed approach, called SwarmClass, outperforms previous ant-based clustering methods and resolve all its drawbacks by the introduction of simple swarm techniques and without the need of complex parameters configuration and prior information on classes’ partition and distribution. Our proposed algorithm uses ants’ segregation behavior to group similar objects together; birds’ moving behavior to control next relative positions for a moving ant; and spiders’ homing behavior to manage ants’ movements when conflicting situations occur.