HDACC: a heuristic density-based ant colony clustering algorithm

Abstract

We present a new heuristic density-based ant colony clustering algorithm (HDACC). Firstly, the device of "memory bank" is proposed, which can bring forth heuristic knowledge guiding an ant to move in the bi-dimensional grid space. Hence the randomness of the ant's motion decreases and algorithm convergence speeds up. In addition, the memory bank makes it possible for every object to be inspected before the algorithm is terminated, which avoids the production of an "unassigned data object". So the classification error rate drops subsequently. Secondly, we proposed a density-based method which permits each ant to "look ahead", which reduces the times of region-inquiry. Consequently, clustering time is saved. We carried out experiments on real data sets and synthetic data sets. The results demonstrated that HDBCSI is a viable and effective clustering algorithm.