Using metaheuristic algorithms to improve k-means clustering: A comparative study

Abstract

Finding the most suitable centroids for k-means clustering is one of the most important criteria for successful clustering operation. We are always looking for the best centroids. Since, clustering problem and finding best centroids are an NP-hard problems, using metaheuristic algorithms can be an appropriate tool to deal with these issues. Many authors have solved this issue with metaheuristic algorithms. Common and popular algorithms have very good solutions. But which of the metaheuristic algorithms really provides the best solution? To answer this question, in this comparative study, ten popular metaheuristic algorithms are compared. The comparisons are performed on synthetic and ten real-world datasets. To find significant differences between the results obtained by algorithms, statistical analysis is used. Comparison results are presented with suitable tables.