Neural network have successfully used forexchange rate forecasting. However, due to a largenumber of parameters to be estimated empirically, it isnot a simple task to select the appropriate neural networkarchitecture for exchange rate forecasting problem.Researchers often overlook the effect of neural networkparameters on the performance of neural networkforecasting. The performance of neural network iscritically dependant on the learning algorithms, thenetwork architecture and the choice of the controlparameters. Even when a suitable setting of parameters(weight) can be found, the ability of the resultingnetwork to generalize the data not seen during learningmay be far from optimal. For these reasons it seemslogical and attractive to apply genetic algorithms.Genetic algorithms may provide a useful tool forautomating the design of neural network. The empiricalresults on foreign exchange rate prediction indicate thatthe proposed hybrid model exhibits effectively improvedaccuracy, when is compared with some other time seriesforecasting models.
展开▼
