Affiliations: [a] Faculty of Electrical and Computer Engineering, University of Science and Technology of Mazandaran, Behshahr, Iran
| [b] Department of Computer Engineering, Jouybar Branch, Islamic Azad University, Jouybar, Iran
Correspondence:
[*]
Corresponding author. Hossein Shirgahi, Department of Computer Engineering, Jouybar Branch, Islamic Azad University, Jouybar, Iran. E-mail: h.shirgahi@jouybariau.ac.ir.
Abstract: This paper proposes a novel evolutionary clustering algorithm and prepares eligible initial centroids for K-Means algorithm by global search approach of efficient hybrid knowledge of swarm intelligence algorithms. Clustering performs data grouping into subsets with common features, So that useful information can be retrieved from them. Swarm intelligence algorithms with evolutionary optimization approach have been very efficient performance in these matters. In this paper, a novel hybrid algorithm called IFAPSO has been proposed which uses swarm hybrid knowledge of Firefly and PSO intelligence algorithms to make an effective data clustering. Performance improvement of PSO and Firefly swarm algorithms and resolve the deficiency of each of them is applied and hybrid of them has been used to benefit from both algorithms into clustering problem effectively. Also this hybrid swarm algorithm overcomes the initial centers’ selection sensitivity and the limitation of local optima in K-Means. We used five benchmarks with several samples and features to evaluate our work. The comparison between proposed methods with the traditional algorithms and previous hybrid methods, suggest that there is more compactness of the resulting clusters and promising accuracy is achieved.
