NEURAL NETWORKS - TOOLS FOR EMPIRICAL ECONOMICS

In recent years there have been major efforts in the development of artificial neural networks (ANN) or connectionist methods to solve problems of pattern recognition, time series forecasting, optimization, modeling of complex systems and related subjects. Important neural architectures are the multilayer-perceptron with backpropagation and the bidirectional associative memory, the BAM-network. A presentation of both methods is given. There are many applications for economic problems, e.g. forecasting of stock exchange rates and of the business cycle. The results are promising but the hopes of the ANN's inventors are not fully satisfied up to now.