Reliable classification using neural networks: a genetic algorithm and backpropagation comparison

Artificial Neural Networks have been shown to have the potential to perform well for classification problems in many different environments, including business, science and engineering. Studies in the literature often report that the artificial neural network dominates traditional statistical techniques for most problems examined. Since a neural network can embed most traditional techniques as special cases we explore why the neural network does not always dominate traditional statistical classification techniques. One reason may be that the majority of these studies rely on a gradient algorithm, typically a variation of backpropagation, to obtain the weights of the model. Although, the limitations of gradient search techniques applied to complex nonlinear optimization problems, such as the artificial neural network, are well known, many researchers still choose to use these methods for network optimization. In this paper, we demonstrate through an intensive Monte Carlo study using real world data sets, that a global search algorithm such as the genetic algorithm overcomes the limitations of these gradient techniques.