Improved k-means clustering with Harmonic-Bee algorithms
Mohammad Babrdel Bonab, Siti Zaiton Mohd Hashim · 2014
Data clustering is one of widely used methods for data mining. The k-means approach is one of the simplest unsupervised learning algorithms that solve the well-known clustering problem. But some hindrances such as the sensitivity to initial values and cluster centers or the risk of trapping in local optimal reduce its best performance. The purpose of kmeans method is minimizing the dissimilarity of observations, from cluster centers. In this paper, a new solution method inspired by harmony search combined with bee algorithm is introduced to improve performance k-means clustering. In this study, harmony and clustering structures are combined to produce harmony clustering. To avoid initial random selection, seed cluster center is considered in primary population as well as bee algorithm has been employed to increase the efficiency of algorithm. The proposed methods have been tested on standard benchmark data sets and also compared to other methods in the literature; it is noted that results show a promising performance leading to better efficiency and capability of the proposed solution.