A Clustering Algorithm Employing Salp Swarm Algorithm and K-Means
Bibi Aamirah Shafaa Emambocus, Muhammed Basheer Jasser, Lim Cher Zet, Samuel-Soma M. Ajibade, Richard T.K. Wong, Hui Na Chua, Ahmad Sahban Rafsanjani
Abstract
Clustering, one of the main types of unsupervised machine learning, consists of grouping data into clusters to discover hidden patterns. Hence it is a crucial machine learning task. The predominant algorithm employed for clustering tasks is the k-means algorithm. However, it has some limitations including being sensitive to the initial centroids. Recently some swarm intelligence algorithms have been noticed to be able to effectively optimize k-means. Hence, in this paper, the Salp Swarm Algorithm (SSA), a recent swarm intelligence with favorable exploration and exploitation capabilities, is employed for optimizing k-means. Specifically, SSA is employed to optimize the initial centroids of k-means to overcome its limitation. The proposed clustering algorithm is applied as part of a movie recommendation system to cluster the users based on their movie preferences. The experimental findings demonstrate that in comparison to the original k-means technique, the proposed clustering algorithm yields superior outcomes as the clustered data by the proposed algorithm has a lower within cluster sum of squares and a higher silhouette score.