Forecasting Exchange Rate Volatility with Linear MA Model and Nonlinear GABP Neural Network

摘要

In order to research RMB exchange rate volatility under exchange rate elastification, this article selects the structure variables about RMB exchange rate volatility to forecast exchange rate volatility by linear moving average model (MA), general back propagation (BP) network and genetic algorithm back propagation (GABP) neural network model respectively. By comparison, we find that, in the lack of flexibility period, month-by-month MA model performs the optimal fitting and forecasting efficiency, along with the exchange rate elastification and liberalization, GABP network model done it best both in volatility value and volatility trend. Exchange rate elastification can deepen the equilibrium relationship between exchange rate and its structure variables, meanwhile, for nonlinear currency fluctuations, nonlinear GABP model could be better choice.