Research On Exchange Rata Forecast And Application Based On MEEMD Combination Model

Nowadays,it is important for us to analyze and forecast the fluctuation of the exchange rate accurately,which will help the government to formulate the financial policy and the enterprises to avoid the foreign exchange risk.In recent years,many models have been widely used in exchange rate prediction,including time series analysis model,neural network model,support vector machine model and other methods.However,owning to a range of disadvantages of the exchange rate series,such as nonlinear,non-stationary,high noise and so on,it is difficult to analyze the characteristics of exchange rate fluctuations and predict the future trend of exchange rate accurately based on the single forecast model or the general combination of forecasting model.The method of multi-scale decomposition has a good characteristics of time-frequency resolution,which can help to decompose the non-stationary sequences into diverse frequency components.In addition,it would choose different forecasting methods and set a multi-scale model according to characteristics and rules of each component sequence.This plays an important role in the analysis of exchange rate fluctuations and the prediction of future trends that has attracted a lot of attentions in the academic circles recently.However,due to the development of multi-scale combination model is still in the initial stage,there are a lot of factors contributing to poor prediction such as inaccurate sequence decomposition,simple reconstruction algorithm and single prediction method.To solve the above problems,it refers to various researches of multi scale and combination model both at home and abroad and constructs a further study of it.Then,in this paper,it establishes a new MEEMD combination model to predict the exchange rate based on a series existing models including improved empirical mode decomposition(MEEMD)model,fuzzy entropy algorithm,Elman neural network model,support vector machine(SVM)model and ARIMA model.The main work of this paper includes the following respects:Firstly,using the MEEMD model and fuzzy entropy algorithm to decompose and reconstruct the exchange rate into high-frequency sequence,which represents the low frequency sequence of major events and representative exchange endogenous institutional and trend factors,analyzes the influence of different factors on the exchange rate.Secondly,according to different wave characteristics of reconstructed items,it selects different combinations of forecasting methods to avoid the problem of low accuracy caused by a single prediction method.The empirical results in experiments show that the forecasting accuracy of MEEMD model is better than other models.Thirdly,a foreign exchange trading strategy based on MEEMD portfolio model is established.Actually,it learns about the settlement series of the exchange rate based on MEEMD portfolio model and designs the adaptive trading threshold that is essential in making decision.Through the analysis of the strategy,we find that the trading strategy has a high rate of return,which can be used in foreign exchange transactions.