An Elman Model Based on GMDH Algorithm for Exchange Rate Forecasting
Qiumin Li, Yixiang Tian, Gaoxun Zhang · International Conference on Management Science and Engineering · 2014
Since the Elman Neural Networks was proposed, it has attracted wide attention. This method has fast convergence and high prediction accuracy. In this study, a new hybrid model that combines the Elman Neural Networks and the group method of data handling (GMDH) is used to forecast the exchange rate. The GMDH algorithm is used for system modeling. Input variables are selected by the external standards. Based on the output of the GMDH algorithm, valid input variables can be used as an input for the Elman Neural Networks for time series prediction. The empirical results show that the new hybrid algorithm is a useful tool.