A combination of objective functions and hybrid Krill herd algorithm for text document clustering analysis. (August 2018)

• Record Type: Journal Article
• Authors: Abualigah, Laith Mohammad; Khader, Ahamad Tajudin; Hanandeh, Essam Said
• Is Part Of: Engineering applications of artificial intelligence. Volume 73(2018)
• Page Start: 111
• Page End: 125
• Publication Date: 2018-08
• Subjects: Combination of objective functions -- Hybridization -- Krill herd algorithm -- K-mean algorithm -- Text document clustering; Engineering -- Data processing -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; Ingénierie -- Informatique -- Périodiques; Intelligence artificielle -- Périodiques; Systèmes experts (Informatique) -- Périodiques; Artificial intelligence; Engineering -- Data processing; Expert systems (Computer science); Periodicals; 620.00285
• DOI: 10.1016/j.engappai.2018.05.003
• Languages: English
• ISSNs: 0952-1976

Abstract

Abstract: Krill herd (KH) algorithm is a novel swarm-based optimization algorithm that imitates krill herding behavior during the searching for foods. It has been successfully used in solving many complex optimization problems. The potency of this algorithm is very high because of its superior performance compared with other optimization algorithms. Hence, the applicability of this algorithm for text document clustering is investigated in this work. Text document clustering refers to the method of clustering an enormous amount of text documents into coherent and dense clusters, where documents in the same cluster are similar. In this paper, a combination of objective functions and hybrid KH algorithm, called, MHKHA, is proposed to solve the text document clustering problem. In this version, the initial solutions of the KH algorithm are inherited from the k -mean clustering algorithm and the clustering decision is based on two combined objective functions. Nine text standard datasets collected from the Laboratory of Computational Intelligence are used to evaluate the performance of the proposed algorithms. Five evaluation measures are employed, namely, accuracy, precision, recall, F -measure, and convergence behavior. The proposed versions of the KH algorithm are compared with other well-known clustering algorithms and other thirteen published algorithms in the literature. The MHKHA obtained the best results for all evaluation measures and datasets used among all the clustering algorithms tested.