A Multi-scale Analysis On The Price Of Exchange Rate Based On EEMD Model

In the context of globalization,China's economy is opening up more and more,and the role of RMB in world economic exchanges is becoming more and more important.The "8.11" exchange rate reform was carried out in 2015,then the RMB exchange rate is more market-oriented.On October 1,2016,RMB joint the SDR currency basket,as a result the RMB is more influenced by the world.It's predictable that Central Banks will take further measures to gradually increase the RMB exchange rate's flexibility.So to reseach the price prediction of the exchange rate market and analysis its risk is very meaningful.For the non-linear,non-stationary and other characteristics of the RMB exchange rate,this paper starts from the angle of multi-scale analysis,combined with the idea of decomposition-reconstruction –integration to use econometric model,machine learning,integrated analysis technology for makeing a comprehensive analysis of the RMB exchange rate data,then to establish a set of integrated prediction model of RMB exchange rate of short-term and long-term data.The concrete steps are as follows: firstly,to use the improved EEMD multi-scale decomposition method to decompose the exchange rate of RMB to the US dollar,then calculate the sample entropy of each decomposition term,and take account of the fluctuation frequency to reconstruct the decomposition term for studying its inherent change mechanism.Next,according to the characteristics of the different fluctuation of the reconfiguration item,to select suitable model for predicting them..At the last,consum the predicting result to get the final result.The paper fount:(1)the fluctuation component of the RMB exchange rate against the dollar including the high frequency components caused by short-term imbalance ?the low frequency components caused by major events?the trend term decided by economic fundamentals,and the low frequency components dominate the fluctuation of RMB exchange rate,followed by the trend,and the effects of high frequency components can be neglected;(2)For long-term data of RMB,the performance of multi-scale prediction is not good,but for the short-term data,compared with the classical time series model and machine learning model,multi-scale EEMD prediction model has higher prediction accuracy and better performance,that's said the structure of component by EEMD decomposition effectively describes the internal operation characteristics of the RMB exchange rate.With the market mechanism in the form of exchange rate is more and more obvious,the derivative RMB exchange rate are developing constantly,and its influence on the development trend and fluctuation of exchange rate itself is moe deeply.Based on the angle of multiscale perspective,use two element EMD algorithm and Copula theory to discuss the relationship between the two marckets from the whole and the micro parts,then apply nonlinear Granger causality test to study the causal relationship between them.The results show that the whole and micro parts of the RMB exchange rate and the RMB NDF have asymmetric,time-varying relationship,and have mutual significant causal relationship.