Analysis And Forecast Of The Relationship Between RMB Daily Exchange Rate And Economic Basic Variables

Exchange rate has always been a key economic index for a country to measure economic and financial fluctuations at home and abroad.It is not only an indicator of economic competition between countries,but also an index of business relations restricted by the value of competitive currencies.In the past,the trend of RMB exchange rate was more affected by policies,but with the many reforms of the exchange rate system and the demand for China to expand its opening to the outside world,the degree of marketization of RMB is bound to deepen,and the influence of policies on the exchange rate is gradually weakening.Therefore,it becomes more and more important to grasp the future exchange rate trend in combination with market factors to provide help for the formulation of exchange rate-related policies and risk behavior decisions.In this paper,according to the daily exchange rate value of RMB and related economic basic variables,several data mining methods are used to analyze and compare the data.The results show that the economic basic variables will have a great impact on the trend of RMB daily exchange rate value.The specific research contents of this paper are as follows:The first chapter introduces in detail the research background,significance and purpose of the exchange rate market,understands the progress of the market-oriented reform of the domestic exchange rate,and then summarizes the current research situation of the exchange rate market at home and abroad.Chapter 2 mainly introduces the data sources,the handling of the abnormal values of the collected data and the statistical analysis of the daily exchange rate value.From October 9,2006 to December 3,2019,excluding the holiday data,the daily data of the effective trading days of US dollar against RMB,POUND against RMB,EURO against RMB and RMB against JAPANESE YEN are selected as response variables.There are twelve characteristic variables in each currency pair,and four kinds of daily exchange rate values are statistically analyzed.Chapter 3 makes further data mining based on time series analysis and feature engineering.Time series analysis shows that the exchange rate value is a time series with trend components,random components and no seasonal components.Through feature correlation analysis,clustering analysis and scatter graph matrix analysis,the internal relationship between features is excavated and shown intuitively,and it is proved that there is correlation between features.Finally,the principal component in feature extraction is used to reduce dimensionality.Find important features,reduce the number of features,and eliminate correlation.Chapter 4 classifies and predicts the rise and fall data based on the feature variables obtained by feature selection.Classified prediction includes naive Bayesian,random forest and GBDT,XGBoost model.Overall,the prediction performance of integrated learning type is the best,and the highest AUC score is 0.869.The experimental results show that the economic basic variables collected in this paper have an impact on the fluctuation of the daily exchange rate,but there is a time difference of 20 working days in the transmission of this influence,which confirms the feasibility of predicting the exchange rate value by the economic basic variables.Chapter 5 makes regression prediction of the data based on the principal component factors obtained by feature extraction.Three kinds of machine learning regression models(linear regression,self-vector regression,gated cycle unit neural network),four kinds of currency pairs and five time step parameters are set for iterative training and regression prediction respectively.The results show that the GRU neural network model has the highest accuracy in the regression prediction of USD/RMB,R2 is 0.9842 and RMSE equals 0.0285,MAE equals 0.0239 MAPE equals 0.0035,and the results show that the step size of regression prediction is best set within 10 trading days,from which we can see that the short-term regression prediction is excellent.Chapter 6 makes a general summary of the previous analysis,enumerates the main research methods and conclusions,analyzes the shortcomings and looks forward to the future development direction.