An efficient robust hyperheuristic clustering algorithm

• Main Author: Bonab, Mohammad Babrdel
• Format: Thesis
• Language: English
• Published: 2016
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://eprints.utm.my/id/eprint/78818/1/MohammadBabrdelBonabPFC2016.pdf; http://eprints.utm.my/id/eprint/78818/; http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:107065
• Institution: Universiti Teknologi Malaysia

Observations on recent research of clustering problems illustrate that most of the approaches used to deal with these problems are based on meta-heuristic and hybrid meta-heuristic to improve the solutions. Hyperheuristic is a set of heuristics, meta- heuristics and high-level search strategies that work on the heuristic search space instead of solution search space. Hyperheuristics techniques have been employed to develop approaches that are more general than optimization search methods and traditional techniques. In the last few years, most studies have focused considerably on the hyperheuristic algorithms to find generalized solutions but highly required robust and efficient solutions. The main idea in this research is to develop techniques that are able to provide an appropriate level of efficiency and high performance to find a class of basic level heuristic over different type of combinatorial optimization problems. Clustering is an unsupervised method in the data mining and pattern recognition. Nevertheless, most of the clustering algorithms are unstable and very sensitive to their input parameters. This study, proposes an efficient and robust hyperheuristic clustering algorithm to find approximate solutions and attempts to generalize the algorithm for different cluster problem domains. Our proposed clustering algorithm has managed to minimize the dissimilarity of all points of a cluster using hyperheuristic method, from the gravity center of the cluster with respect to capacity constraints in each cluster. The algorithm of hyperheuristic has emerged from pool of heuristic techniques. Mapping between solution spaces is one of the powerful and prevalent techniques in optimization domains. Most of the existing algorithms work directly with solution spaces where in some cases is very difficult and is sometime impossible due to the dynamic behavior of data and algorithm. By mapping the heuristic space into solution spaces, it would be possible to make easy decision to solve clustering problems. The proposed hyperheuristic clustering algorithm performs four major components including selection, decision, admission and hybrid metaheuristic algorithm. The intensive experiments have proven that the proposed algorithm has successfully produced robust and efficient clustering results.