Enhancing Classifiers through Neural Network Ensembles

Artificial neural networks are known to have strong generalization abilities, but they entirely lack comprehensibility, due to their connectionist nature. Neural network ensembles augment this characteristic, making them less appealing in domains where comprehensibility is as important as accuracy. This paper presents the implementation of a new system based on a method for combining ensembles of neural networks with symbolic learners. The focus is on enhancing the symbolic classifiers by using a neural network ensemble as a pre-processing step for them. The results obtained during the evaluations on the new system have confirmed that the approach is suitable for enhancing the performance of symbolic classifiers.