Chaotic time series modeling with optimum neural network architecture

A neural network approach for modeling and predictions on chaotic time series is presented. The main aim is to reduce the size and complexity of the network and use the least number of weights and nodes for any predictive mapping. The problem of selecting the number of input and hidden nodes is studied by the predictive minimum description length principle. We discuss comparatively the performance of neural networks and conventional methods in predicting chaotic time series. The neural network is found to yield better predictions than an optimum ARMA model.