CURRENCY CRISIS FORECASTING WITH GENERAL REGRESSION NEURAL NETWORKS

The main purpose of this study is to devise a general regression neural network (GRNN)-based currency crisis forecasting model for Southeast Asian economies based upon the disastrous 1997–1998 currency crisis experience. For this some typical indicators of currency exchange rates volatility are first chosen, then these indicators are input into GRNN for training, and finally the trained GRNN is used for future crisis prediction. To verify the effectiveness of the proposed currency crisis forecasting approach, four typical Southeast Asian currencies, Indonesian rupiah, Philippine peso, Singapore dollar and Thai baht, are selected. Meantime we compare its performance with those of other forecasting methods to evaluate the forecasting ability of the proposed approach. Empirical results obtained reveal that the proposed currency crisis forecasting model has a surprisingly high degree of accuracy in judging the currency crisis level of each country in specified time period, implying that our proposed approach can be used as a feasible currency crisis early-warning system to predict currency crisis level for other countries around the world.