Swarm ascending: Swarm intelligence-based exemplar group detection for robust clustering

An exemplar is a representative observation for each cluster. Exemplar-based clustering algorithms, which find the exemplars and assign data points to the nearest exemplar, have exhibited promising performance. However, the single- and multi-exemplar methods become inadequate for clustering data points with nonlinear and local patterns because one exemplar (or a set of sparse exemplars for a nonlinear cluster) is insufficient to represent the cluster. In this paper, we propose a swarm intelligence-based exemplar group detection method that ascends data points to local high-density points and groups the merged points. The proposed method is robust to nonlinear and local patterns because it detects the intrinsic structure of each cluster more sufficiently than sparse exemplars. We use simulation and real-world data to demonstrate the usefulness of the proposed method by comparing it to existing methods in terms of clustering accuracy. The comparison results demonstrate that the proposed method outperforms the alternatives.