Exchange rate prediction method based on ARIMA-HPSO-Elman combined model with SSA： Based on the central parity rate data of USD/CNY

Exchange rate has the characteristics of both linear and non-linear mixed behavior. Single linear model and non-linear model are not perfect for forecasting exchange rate.Here the central parity rate series of USD/CNY exchange rate was studied. Firstly, the SSA method was used to denoise the exchange rate series, and ARIMA model was established to fit and predict the reconstructed exchange rate series to extract the linear components of the original exchange rate series. Secondly, the residual part was modeled and predicted by Elman neural network optimized by hybrid particle swarm optimization algorithm based on crossover and mutation. The sum of the results was the predicted value of the original exchange rate series. Empirical results show that CNY exchange rate fluctuation has the characteristics of periodic oscillation. On the 30-day forecast outside the sample of exchange rate series, the performance of the combination model based on SSA method is better than that of the single model and the combination model without SSA method in the short term.