Research On The Linkage Relationship Between Short-term Exchange Rate And Gold Price Based On Intelligent Computin

Since the reform and opening up,China has maintained high-speed economic growth for 40 years,and made great achievements in the financial industry.Especially in the past decade,China attaches great importance to the financial work,and the international competitiveness and influence of the financial industry have been greatly increased.In the future,China will continue to balance financial openness and security,and make every effort to build a financial management system suitable for high-level opening up.The report of the 20 th National Congress of the Communist Party of China clearly pointed out that to "strengthen and improve modern financial supervision,strengthen the financial stability guarantee system",strengthen the supervision of the financial market,and maintain the stability of the financial market is one of the primary tasks of deepening the financial reform and opening up,and is an important guarantee for maintaining the stable operation of the financial market and the national economic and financial security.Modern financial markets not only have the characteristics of dissemination,infectivity and risk spillover,but also have strong joint mobility and sharing.Research on the linkage mechanism and transmission mechanism of financial markets is of positive significance for accurately grasping the dynamic change law of financial markets and establishing and improving the regulatory framework of modern financial markets that combines macro-prudential coordination and micro-supervision.This dissertation studies the linkage between the exchange rate market and the gold market,an important part of the financial market,mainly including steady-state analysis and transient analysis.With the steady progress of market-oriented reform of China’s exchange rate,the market-oriented mechanism of RMB exchange rate has been gradually improved,the flexibility of exchange rate has been continuously enhanced,and two-way fluctuations have become normal.The RMB exchange rate has remained basically stable at a reasonable and balanced level,but the uncertainty in the exchange rate market and the interaction of various financial factors have formed various financial transmission mechanisms.The accurate grasp of the exchange rate market plays a very important role in maintaining the stability of the financial market.On the other hand,because of the three attributes of commodity,currency and finance,gold is not only an important precious metal that actively interacts with other global asset markets and fully priced,but also an important embodiment of currency value.It has the function of resisting currency devaluation in the long term.The short-term exchange rate and gold price have a very close relationship with each other.Therefore,the study of the linkage between short-term exchange rate and gold price in the context of modern financial markets has important theoretical and practical significance for financial policy designators,enterprises and individual investors.Different from the widely used research methods of statistics,this dissertation studies the linkage between short-term exchange rate and gold price based on intelligent computing,and organically combines intelligent computing with traditional statistical methods,which not only serves as the exploration of new methods of time series analysis,but also provides new analytical means for the linkage between exchange rate and gold price.In this dissertation,the interaction between short-term exchange rate and gold price is firstly analyzed.Based on the extensive research on the effect of exchange rate on gold price in the existing literature,the effect of gold price on short-term exchange rate is focused on.Secondly,in view of the nonlinear relationship between the dollar index and the gold price,two data-driven neural learning methods for nonlinear error correction are proposed by combining the neural network with the error correction model(ECM),so as to improve the goodness of fit between time series.Thirdly,combining adaptive Fourier decomposition(AFD)with frequency domain impulse response analysis,the adaptive impulse response model with energy resolution is discussed,and the linkage between the US dollar index and the gold price series is characterized and the impulse response analysis is carried out.Finally,on the basis of the above steady-state analysis of the linkage between gold price and exchange rate,aiming at the time variability of time series in the case of drastic market fluctuations after the occurrence of risk events,the directed acyclic graph technique is combined with the generalized autoregressive conditional heteroscedasticity model with dynamic conditional correlation,and the transient analysis of the risk transmission mechanism between exchange rate and gold price is carried out.The combination of intelligent computing and traditional statistical methods explored in this dissertation,on the premise of meeting the requirements of statistical testing and the explanatory nature of variables,expands the traditional statistical methods in dealing with the nonlinearity and time variability of time series and the adaptability of the proposed algorithm,and provides a new analytical approach for the analysis of modern financial time series.The main research contents and achievements are as follows:1)The ARMAX-GARCHX exchange rate model based on gold parity is proposed.Aiming at the effectof gold price on exchange rate,the ratio of the price of "Shanghai gold" denominated in RMB and the price of "London gold" denominated in USD is studied as the gold parity RMB exchange rate,and the effectof gold parity on the trend of USD/RMB onshore exchange rate is discussed.With the gold parity as an exogenous variable,the ARMAX-GARCHX model of short-term exchange rate is established,and the Kupiec test is used to verify the improvement of the prediction accuracy of the daily frequency return distribution of exchange rate,and the gold parity is quantitatively mapped into the analysis of exchange rate.Based on daily frequency data,the empirical analysis was carried out.The results show that gold parity is one of the factors affecting the short-term exchange rate trend.It has positive significance for improving the accuracy of short-term exchange rate forecast,policy formulation and risk prevention,and has important reference value for the exchange rate marketization,the financial risk prevention in the process of RMB internationalization and the daily exchange rate risk management of enterprises.2)Two nonlinear error correction learning models based on neural network are proposed to describe the nonlinear relationship between dollar index and gold price.In view of the nonlinear characteristics of time series and the limitation that the error correction model still uses linear regression,an error correction learning model integrating long and short time memory recurrent neural network(LSTM)is proposed.This model uses the nonlinear characteristics and long memory of neural network to improve the nonlinear expression ability of time series,and also improves the economic connotationof variables in the neural network.The nonlinear error correction learning model was used to analyze the nonlinear linkage between the US dollar index and the gold price year by year,and its goodness of fit is significantly higher than the traditional error correction model.At the same time,a hybrid neural network learning nonlinear error correction model is constructed by combining linear recursive neural network(RNN)with multilayer error back propagation(BP)network.The network learning algorithm is given by using gradient descent method and error back propagation.Based on the principle of data-driven,all network parameters can be obtained through network learning and training.The proposed nonlinear ECM neural learning method was verified by using the daily frequency data of the gold price and the dollar index,and the results were compared by the likelihood ratio Chi-square test.The simulation results show that the proposed data-driven nonlinear error correction neural learning method can statistical significantly improve the goodness of fit of the relationship between complex nonlinear time series.3)An adaptive impulse response model of time series based on adaptive Fourier decomposition(AFD)is proposed and applied to the impact analysis of impulse response of time series.The relationship between linkage and impulse response is described from the perspective of "energy resolution".According to the observed time series,the frequency point of the impulse response function is estimated,and the approximate expression of the impulse response function in the frequency domain is given;The approximate impulse response function is AFD step by step,and the poles and coefficients of the basis function are adaptively obtained according to the energy maximization criterion.When goodness of fit is greater than a certain threshold,all decomposition levels are determined and an adaptive impulse response model is established.Each decomposition level describes the energy-based resolution characteristics of the impulse response model.The impact response was analyzed year by year based on the gold price yield and the daily frequency yield of the US dollar index.The results show that the proposed adaptive impulse response model based on AFD can describe the sustained effect of gold price shock on the dollar index,and further measure the flatness of the correlation between the two time series.Decomposing the impulse response to each decomposition level can reveal the consistency,convergence and sufficiency of the impulse response model.The adaptive data-driven model based on AFD provides a new method for time-frequency analysis of time series.4)A directed acyclic graph(DAG-DCC-GARCH)model based on generalized autoregressive conditional heteroscedasticity with dynamic conditional correlation is proposed to describe the transient risk transmission mechanism between markets,especially the changes of risk transmission mechanism in the case of violent market fluctuations after the occurrence of risk events.DCC-GARCH is used to calculate the dynamic correlation coefficient matrix,and a time-varying directed acyclic graph(DAG)reflecting the causal relationship between the daily frequency and the same period is constructed.Bootstrap method is used to test the significance of correlation coefficient and partial correlation coefficient.The obtained daily instantaneous directed acyclic graph depicts the instantaneous risk transmission mechanism between capital markets.The empirical results of the five major capital markets in China and the United States separately show that the stock market and gold market in China and the US are the main instantaneous risk spillovers.The higher the risk level of the overall market,the lower the risk spillover intensity between the capital markets,which reflects that the two countries have a certain effect in systemic risk prevention.The instantaneous risk transmission intensity of the Chinese market is higher than that of the American market,and increases year by year.At the same time,the instantaneous risk transmission attribute of the American market is more stable than that of the Chinese market.In the study of the event impact represented by the early outbreak of COVID-19,the overall instantaneous risk transmission capacity of the two countries gradually decreased after the occurrence of the event,and the stock market showed relatively stronger risk spillover capacity.According to the research on the risk transmission in the transnational markets of gold and foreign exchange markets in China and the United States,the risk spillover capacity of gold is stronger than that of foreign exchange markets,especially Shanghai Gold has the strongest risk spillover capacity in the four markets.Volatility from the gold market often has an impact on the risk transmission mode,while volatility from the foreign exchange market has a limited impact on the transmission mode of the four market events.