Hybrid Metaheuristic Algorithm for Clustering
Olayinka Idowu Oduntan, Parimala Thulasiraman · 2018
Clustering involves grouping a collection of data objects into meaningful or useful categories such that objects within the same category are similar to one another while objects in different categories are dissimilar. Clustering is a challenging problem with diverse practical applications that span multiple research domains. A review of existing literature shows that there are many diverse clustering algorithms for different problem domains. Also, many popular optimization heuristics and metaheuristics have been adapted to create clustering algorithms, but these algorithms typically inherit the limitations of the underlying heuristics or metaheuristics. An evolving trend in metaheuristic algorithm design is to combine concepts and/or components from multiple algorithms to tackle difficult optimization problems such as clustering. In this research, we explore the possibility of harnessing the strengths of multiple metaheuristic algorithms to tackle the clustering problem. We propose a hybrid metaheuristic algorithm for clustering that combinesant brood sorting (a nature-inspired clustering technique) with tabu search (a metaheuristic that uses search history and dynamic neighborhood strategies to uncover global optimal solution). This is a new hybrid metaheuristic approach to clustering with emphasis on flexibility and less specificity.