Data Clustering using Two-Stage Eagle Strategy Based on Slime Mould Algorithm

Abstract

Dataclustering is considered an important component of data mining which aims tosplit a given dataset into disjoint groups having the same similarities. Thedeveloped techniques for clustering have some challenges to cluster entities incomplex search space and most of them aim to maximize the sum of inter-clusterdistances and minimize the sum of intra-cluster distances. This objectivefunction is nonlinear and hard to optimize especially for complex search space.Metaheuristics are becoming a trend for solving this task thanks to theirpromising results. In this study, the eagle strategy is used to take advantageof the exploration provided by Levy Flight (LF) and the exploitation strengthof the Slime Mould Algorithm (SMA) to solve the clustering problem. The SMAalgorithm is an efficient technique for solving complex optimization problemswhich has a high exploitation competence. On the other hand, LF tends to havegood exploratory behavior. Our strategy exploits these advantages in a balancedway and through well-designed rounds to ensure the optimality of the clusteringsolutions. The proposed method is computationally efficient and inexpensive. Italso achieves high accuracy in terms of average, worst, best, and the sum ofintra-cluster distance. The method is also evaluated according to the speed ofconvergence and using statistical tests, namely Wilcoxon. The obtained resultsare compared with seven benchmarked metaheuristics, namely Grey Wolf Optimizer(GWO), Slime Mould Algorithm (SMA), Whale Optimization Algorithm (WOA), HarrisHawks Optimization (HHO), Sine Cosine Algorithm (SCA), Multi-Verse Optimizer (MVO)and Genetic Algorithm (GA) using eighteen datasets of shapes and UCIrepositories.