Efficient computation of fitness function for evolutionary clustering

Abstract

Evolutionary algorithms (EAs) are random search heuristics which can solve various optimization problems. There are plenty of papers describing different approaches developed to apply evolutionary algorithms to the clustering problem, although none of them addressed the problem of fitness function computation. In clustering, many clustering validity indices exist that are designed to evaluate quality of resulting points partition. It is hard to use them as a fitness function due to their computational complexity. In this paper, we propose an efficient method for iterative computation of clustering validity indices which makes application of the EAs to this problem much more appropriate than it was before.