EFuNNs Ensembles Construction Using a Clustering Method and a Coevolutionary Genetic Algorithm

Abstract

This paper introduces a new approach to construct neural network ensembles called clustering and co-evolution to construct neural network ensemble (CONE). This approach was used to create and optimize the parameters of a particular type of evolving fuzzy neural networks (EFuNNs) ensemble. Experimental results on four benchmark databases show that the CONE generates EFuNNs ensembles with accuracy either better or equal to the accuracy of single EFuNNs generated using a genetic algorithm. Besides, the execution time of CONE to generate EFuNNs ensembles is lower than the execution time of the genetic algorithm to produce single EFuNNs.