Automatic clustering based on invasive weed optimization algorithm

Abstract

In this article, an evolutionary metaheuristic algorithm known as the Invasive Weed Optimization (IWO) is applied for automatically partitioning a dataset without any prior information about the number of naturally occurring groups in the data. The fitness function used in the genetic algorithm is a cluster validity index. Depending on the results of this index IWO returns the segmented dataset along with the appropriate number of divisions. The proficiency of this algorithm is compared to variable string length genetic algorithm with point symmetry based distance clustering(VGAPS-clustering), variable string length Genetic K-means algorithm(GCUK-clustering) and a weighted sum validity function based hybrid niching genetic algorithm(HNGA-clustering) and is denoted for the nine artificial datasets and four real life datasets. © 2011 Springer-Verlag.