A CLASSIFICATION APPROACH USING MULTILAYERED NEURAL NETWORKS

There has been an increasing interest in the applicability of neural n etworks in disparate domains. In this paper, we describe the use of mu lti-layered perceptrons, a type of neural-network topology, for financ ial classification problems, with promising results. Back-propagation, which is the learning algorithm most often used in multi-layered perc eptrons, however, is inherently an inefficient search procedure. We pr esent improved procedures which have much better convergence propertie s. Using several financial classification applications as examples, we show the efficacy of using multilayered perceptrons with improved lea rning algorithms. The modified learning algorithms have better perform ance, in terms of classification/prediction accuracies, than the metho ds previously used in the literature, such as probit analysis and simi larity-based learning techniques.