Data analysis by combining the modified k-means and imperialist competitive algorithm

Abstract

Data Clustering is one of the most used methods of data mining. The k-means Clustering Approach is one of the main algorithms in the literature of Pattern Recognition and Data Machine Learning which it very popular because of its simple application and high operational speed. But some obstacles such as the adherence of results to initial cluster centers or the risk of getting trapped into local optimality hinders its performance. In this paper, inspired by the Imperialist Competitive Algorithm based on the k-means method, a new approach is developed, in which cluster centers are selected and computed appropriately. The Imperialist Competitive Algorithm (ICA) is a method in the field of evolutionary computations, trying to find the optimum solution for diverse optimization problems. The underlying traits of this algorithm are taken from the evolutionary process of social, economic and political development of countries so that by partly mathematical modeling of this process some operators are obtained in regular algorithmic forms. The investigated results of the suggested approach over using standard data sets and comparing it with alternative methods in the literature reveals out that the proposed algorithm outperforms the k-means algorithm and other candidate algorithms in the pool