A novel spiral optimization for clustering

Abstract

Because most traditional search methods are unable to satisfy the current needs of data mining, finding a high performance search method for data mining has gradually become a critical issue. The spiral optimization (SO) is a promising search algorithm designed to emulate the natural phenomena, such as swirl and low pressure, to find the solutions of optimization problems within an acceptable computation time. In this paper, a novel SO is presented to solve the clustering problem. Unlike the original SO, which rotates the points around the elitist center iteratively, the proposed algorithm, called distributed spiral optimization (dSO), splits the population into several subpopulations so as to increase the diversity of search to further improve the clustering result. The k-means and oscillation methods are also used to enhance the efficacy of dSO. To evaluate the performance of the proposed algorithm, we apply it to the clustering problem and compare the results it found with those of the spiral optimization and genetic k-means algorithm. The results show that the proposed algorithm is quite promising. © Springer-Verlag Berlin Heidelberg 2014.