Data clustering by ant colony on a digraph

An adaptive data clustering algorithm based on ant colony (ant-cluster) is presented. Enlightened by the self-organizing behavior of ant society, we assign acceptance rates on the directed edges of a pheromone digraph in ant-cluster system. The pheromone on the edges of the digraph is adaptively updated by the ants passing it. Some edges with less pheromone are progressively removed under a list of certain thresholds in the process. Strong connected components of the final digraph are extracted as clusters. The performance of ant-cluster is compared with classical K-means clustering algorithm and ACO clustering algorithm LF in terms of clustering quality and efficiency on several real datasets and clustering benchmarks. Experimental results indicate that the ant-cluster is able to find clusters faster with better clustering quality and is easier to implement than K-means and LF.