An economic forecasting system based on recurrent neural networks
Jian Chen, Dong Yu Xu · 2002
Many methods for economic forecasting have been developed, and most of them are based on statistical techniques. To increase the capability of describing economic processes and the accuracy of economic forecasting, some efforts for applying artificial neural networks to economic forecasting have been made, which are mostly based on multilayered feedforward network (MFN). Compared with MFN, recurrent neural networks have the ability to consider the historical deviation for further modification of the forecasting model. In this paper, a forecasting system based on recurrent neural networks is presented for economic forecasting and business cycle prediction. To speed up the training process, an improved backpropagation algorithm is embedded in the system. This system has some attractive properties and is now being used by the city government for supporting their policy making.