Abstract
Data clustering is commonly employed in many disciplines. The aim of clustering is to partition a set of data into clusters, in which objects within the same cluster are similar and dissimilar to other objects that belong to different clusters. K-means (KM) and K-harmonic-means (KHM) are two common and fundamental clustering methods because of their simplicity and efficiency. However, both of them suffer from some problems. This study presents two novel algorithms based on simplified swarm optimization to deal with the drawbacks of KM and KHM, respectively. In addition, with the advance of internet and information technologies, the data size is increasing explosively and many existing clustering approaches including KM and KHM are inefficiency for dealing with the large-size problem. For that, we propose a clustering framework by exploring the connection between principle component analysis and a novel random sampling technique into a procedure to increase the scalability of the proposed clustering algorithm. To empirically evaluate the performance of the proposed methods, all experiments are examined using real-world datasets, and corresponding results are compared with recent works in the literature.