Title: Clustering using modified harmony search algorithm
Authors: Vijay Kumar; Jitender Kumar Chhabra; Dinesh Kumar
Department of Computer Science and Engineering, JCDM College of Engineering, Sirsa-125055, Haryana, India
Department of Computer Engineering, National Institute of Technology, Kurukshetra-136119, Haryana, India
Department of Computer Science and Engineering, Guru Jambheshwar University of Science and Technology, Hisar-125001, Haryana, India
Abstract: Metaheuristic techniques are being successfully used as optimisation methods in various application areas. This paper presents modification in one such metaheuristic called as harmony search (HS) that is inspired from music improvisation process. The two parameters, harmony memory consideration rate (HMCR) and pitch adjusting rate (PAR), in HS play important role in improvisation of new harmony. Instead of keeping the parameters fixed, as reported in many existing algorithms, these are being allowed to change dynamically during the process of improvisation in the proposed algorithm. This paper further examines the effect on the results when K-means is initialised with solution returned by the proposed algorithm. The effect of harmony memory size has also been investigated on proposed approach. The experiments are performed for data clustering on nine benchmark datasets. The clustering performance of proposed algorithm is compared with K-means, fuzzy C-means, genetic algorithm, and four recently proposed variants of HS. The results are encouraging and demonstrate that the proposed algorithm provides much better values in terms of precision, recall, G-measure, inter-cluster and intra-cluster distances.
Keywords: metaheuristics; modified harmony search; K-means clustering; harmony memory size; data clustering; fuzzy C-means clustering; genetic algorithms; harmony memory consideration rate; pitch adjusting rate.
Int. J. of Computational Intelligence Studies, 2014 Vol.3, No.2/3, pp.113 - 133
Available online: 11 Jun 2014