Electromagnetism based K-means clustering for big data
Date
Authors
ORCID
Journal Title
Journal ISSN
Volume Title
Publisher
DOI
Abstract
Over the past few years, Nature has been the source of inspiration for many proposed successful algorithms. This paper proposes a new nature-inspired K-means clustering algorithm which is based on the concept of Electromagnetism. The proposed algorithm starts by initializing a set of particles and later in the second step, the best particle among them is chosen based on the fitness function. After choosing the best particle, an objective function value is calculated for each particle which is initialized. Then the force and movement are calculated for each particle except for the current best particle. This way, the algorithm at each iteration searches for a local best particle and then calculates objective function values. Due to this reason, the position of the initialized particles also changes. Algorithm terminates when it reaches the maximum iterations or when the change in Within Set Sum of Squared Error (WSSSE) is less than 0.0001. The detailed explanation of this algorithm is presented. From the results, Electromagnetism based K-means provides better accuracy when compared to K-means clustering. This can be seen from the results section.