Acomparative Study of the Predictive Accuracy of Multilayer Perceptron Networks Versus Simple Recurrent Networks for Selecting Forcasting Methods

摘要

Forecasting is an important and necessary activity for ally types of organizations. Selecting a forecasting model is a complex and time-consuming task. This research describes the use of two Artificial Neural Network topologies namely, Multilayer Perceptron and Simple Recurrent Networks to predict the msot accurate forecasting model for any given time series. Econometric time series datasets were used to train and test the ANNs. The results of this experiemnt, which appear promising are presented together with guideliens for its practical application. Potential benefits include dramatic reductions in the effort and cost of forecasting; the provision of assistance to specialist forecasters; and increases in forecasting accuracy.