Neural Network Ensembles for Classification Problems Using Multiobjective Genetic Algorithms

In this work a Multiobjective Genetic Algorithm is developed in order to obtain an appropriate ensemble of neural networks. The algorithm does not use any back-propagation method. Furthermore, it considers directly the classification error instead of the mean square error. To obtain the multiobjective environment, the training pattern set is divided into subsets such that each one has its own error function and then, all the error functions are considered simultaneously. The proposed algorithm is found to be competitive with other current methods in the literature.