Variable neighborhood search algorithm for k-means clustering

Abstract

We propose new algorithms of Greedy Heuristic Method for solving the classical problem of cluster analysis, k-Means, which allows us to obtain results with better objective function values in comparison with known algorithms such as k-Means and j-Means. Their comparative efficiency is proved by experiment on various data sets including multidimensional data of non-destructive rejection tests of electronic components for the space industry.