An Improved Clustering Algorithm via Salp Swarm Algorithm

Abstract:
Clustering is an unsupervised learning method and a common statistical data analysis technique. The purpose of data clustering is to organize similar objects into one cluster optimally. The key for the clustering problem is to find a suitable cluster center. Currently, many optimization algorithms are used to search cluster centers. In this work, we develop a new Salp Swarm Algorithm (SSA) for clustering problems. SSA is an algorithm that simulates the behavior of salparia populations to achieve optimum. Then, the SSA-clustering is tested on four UCI datasets and compared with other 8 popular algorithms. The experimental results show that the intra-cluster distance obtained by this optimizer is smaller than any other algorithms. We can conclude the SSA clustering can handle clustering problems more effectively.