Adaptive Ant Clustering Algorithm with Pheromone

In the midst of data mining tasks, clustering algorithms received special attention, especially when these techniques are bio-inspired and while they use special methods which improve a learning process during clusterization. Most promising among them are ant-based approaches. The process of clustering with colony of virtual ants is emerging and can be an alternative, when the data is complicated. Clustering, based on ant’s behavior, was first introduced by Deneubourg et al. in 1991 and this classical proposition still requires investigation to improve stability, scalability and convergence of speed. This investigations will show that we can create a mature tool for clustering. The aim of this research was to examine the execution of a new Ant Clustering Algorithm with a modified scheme of ants’ perception and an incorporation of pheromone matrices. To assess the performance of our proposition, certain amount of widely known benchmark data sets were used. Empirical study of our approach shows that the adACA performs well when the pheromone matrices influence the behavior of clustering ants and leads to better results.