Certain investigations on data clustering using hybrid metaheuristic algorithms

Abstract

Clustering procedures are very useful in different areas such as medical image analysis psychology pattern recognition information retrieval climate analysis business biology intrusion detection system and robotics The main objective of data clustering is to separate or group a large set of available data into meaningful clusters which maximize the intra cluster homogeneity and inter cluster heterogeneity Intra cluster homogeneity defines the degree of similarity between the elements which present in the cluster Clusters are internally homogeneous and it differs from the other clusters which are defined as inter cluster heterogeneity Hybrid metaheuristic algorithm is a combination of more than one algorithm in order to increase the quality of a solution Two main components which justify the quality of the solution are exploration and exploitation Each algorithm has its individual capacity and it may concentrate on either exploration or exploitation When we hybrid more than one algorithm it will help to improve the quality of the solution by focusing on both exploration and exploitation And one more important property is the gap between exploration and exploitation If the gap is very high then it will lead the poor solution On the other hand if the gap is very small in nature then it will yield good quality of the solution Proposed algorithms are developed based on three metaheuristic algorithms such as Ant Lion Optimization Raven Roosting Optimization and Mouth Brooding Fish Optimization algorithms These algorithms are included in intermediate solution generation part of the hybrid methodology The overall result shows that proposed hybrid MBF algorithm provides optimal solution in minimum number of iterations newline