The forecast of USD/CNY exchange rate based on the Elman neural network with volatility updating

摘要

This paper proposes a procedure for using an EGARCH-M model in conjunction with Elman neural network when estimating the return series of USD/CNY exchange rate. Data analysis shows that the returns of USD/CNY exchange rate are fat-tails and have volatility clusters. The EGARCH-M model captures the asymmetry in volatility and the correlation between the return and its past volatility, and the Elman neural network is applied to fit the nonlinear characteristic of the return series. By comparing the evaluation index of different models, it proves that the approach of Elman neural network with volatility updating derives a satisfactory empirical result. And the forecast of the return series illustrates the excellent performance of the model in the end.