Homogeneous Ensemble Selection through Hierarchical Clustering with a Modified Artificial Fish Swarm Algorithm

In the pattern recognition field, ensembles of classifiers have been proposed as a method to overcome the natural limitations of single classifiers, and to increase the accuracy of the system. Previous studies show that ensembles of classifiers need to have accurate classifiers that have different knowledge for the same problem. In this paper, we propose an ensemble selection technique for single layer neural networks trained by the Extreme Learning Machine algorithm based on the Artificial Fish Swarm Algorithm. The ensembles are grouped based on information on the fish population using a hierarchical cluster algorithm. Experimental results show that the proposed method achieve better generalization performance than best model produced by the modified optimization technique presented in real benchmark datasets.