An evolutionary approach to the specification of high performing backpropagation Neural Networks

Genetic Algorithms and Neural Networks have been combined in various ways in an effort to develop powerful tools for problem solving. This work presents a two stage process for the specification of high performing backpropagation neural networks for four commonly used real world databases. GAs are used to evolve a set of potential neural network architectures and operating parameters. The best networks are then used for further training and testing, to determine an optimum starting seed and number of training passes. In order to test and compare networks developed in this way, an exhaustive set of experiments using the machine learning induction algorithm, C4.5 were also carried out on the four databases. These results, together with some previously published benchmark data and BPNN results, were compared with those obtained from the networks developed using our method. Our results compare more than favourably with the C4.5, Holte's 1R* benchmark and previously published BPNN results across the four databases. It is hoped that this work may lead to the development of an integrated tool for use in the solution of difficult real world classification problems.