A Combined Approach Based on K-Means and Modified Electromagnetism-Like Mechanism for Data Clustering

Clustering is one of the useful methods in many scientific fields. It is a classification process to group data in specific clusters based on their similarities. Many heuristic and meta-heuristic algorithms have been successfully applied in the literature to solve clustering problems. Among them, the K-means is one of the best due to its simplicity and computational efficiency. However, it suffers from several drawbacks, the most significant of which is its dependency on the initial state that leads to trapping in local optima. In this paper, the K-means method is combined with a modified electromagnetism-like mechanism (MEM) algorithm to develop a new algorithm called K-MEM in order to avoid trapping in local optima. In addition, two modifications are made in this paper to improve the performance of the EM algorithm. First, a modified local search procedure is adopted to improve searching. Second, an elitism approach is imported to improve the moving procedure. The proposed algorithm is tested on four standard datasets chosen from the UCI Machine Learning repository and several artificial datasets, where its performance is compared with those of EM, MEM, K-means, combination of K-means and EM (K-EM), genetic algorithm (GA), simulated annealing (SA), ant colony optimization (ACO), honey bee mating optimization (HBMO) and particle swarm optimization (PSO). The results illustrate that the proposed K-MEM algorithm has a good performance to find desired results.