Effect of Harmony Search Parameters’ Variation in Clustering

Abstract

Harmony search (HS) is a relatively new meta-heuristic optimization method, which is based on the concept of music improvisation. This paper depicts the impact of constant parameters such as Harmony Memory Consideration Rate and Pitch Adjusting Rate, and presents an approach for parameter tuning. It presents modifications in existing harmony search, by choosing appropriate values of these two parameters and allows them to change dynamically during the process of improvisation. The proposed algorithm has been evaluated for data clustering on five benchmark datasets. The clustering performance of proposed algorithm is compared with K-Means, Genetic algorithm, HS and improved version of HS. Experimental results reveal that proposed algorithm provides better results than the above said techniques in terms of precision, recall, G-Measure, inter-cluster and intra-cluster distance.