Exploring k-PSO Algorithm for Clustering

Cluster analysis is a very popular approach to fully automatic search for patterns, data concepts, groups and clusters. It simplifies data representations  and thus plays an important role in the process of knowledge acquisition. Data mining tasks require fast and accurate partition of data with many attributes. This requires new approach, which could deal better with these features. Methods based on the swarm intelligence present such approach to the cluster analysis. This article is a brief introduction to the optimization algorithms inspired by the natural world. It shows how these algorithms can be used in the cluster analysis and describes several up-to-date hybrid techniques combining PSO and k-means. Moreover, conceptually new hybrid algorithm based on the PSO and k-means is introduced and its efficiency and robustness are compared to the other algorithms using several datasets.