Partial Correlation Analysis Of The International Status Of The US Dollar And Its Trend Prediction Model Selection

After the establishment of the Bretton Woods system in 1944,the US dollar became the international central currency.Although its current status has declined,it still dominates compared with other currencies.However,as of February 11,2023,the total federal debt of the United States has reached $ 31.54 trillion,exceeding the statutory debt ceiling of $ 31.4 trillion,and the total federal debt of the United States as a share of US GDP has exceeded 120 %.The security of holding US debt is worrisome,which in turn has shaken the international community ’s confidence in the international status of the dollar.Given that China currently holds a large amount of US dollar assets,including US debt,accurately judging the international status of the US dollar is of great significance for the country to safeguard the security of US dollar assets.Based on this,this paper analyzes the partial correlation of the international status of the US dollar,and selects the prediction model of the US dollar trend,in order to predict the future trend of the international status of the US dollar more accurately.This paper introduces the Arima-R-Vine-Copula model to study the partial correlation of the international status of the US dollar.When studying the relationship between the US dollar index and the US federal government bond yield,the proportion of the US fiscal deficit,the US economic growth rate,the US current account balance,the US capital account balance and the proportion of the US dollar in foreign exchange reserves are regarded as control variables.The results show that there is a negative correlation between the US dollar index and the US federal bond yield.According to the rank correlation coefficient matrix,the correlation coefficient between the US dollar index and the US federal bond yield is-0.09.From the Arima-R-Vine-Copula with a fixed center point,it can be concluded that as the control variables are introduced one by one,the correlation coefficient between the US dollar index and the US federal bond yield changes.When unconditional,the correlation coefficient between the US dollar index and the US federal bond yield is-0.10,indicating that when the US federal bond yield rises,the international status of the US dollar declines,and vice versa;when the proportion of the US fiscal deficit is used as the control variable,the correlation coefficient between the US dollar index and the US federal bond yield does not change significantly;when the US fiscal deficit ratio and the US economic growth rate are both used as control variables,the correlation coefficient between the US dollar index and the US federal bond yield increases slightly,indicating that when the US federal bond yield rises,the decline in the US dollar ’s international status is enhanced,and vice versa;when the proportion of the U.S.fiscal deficit,the U.S.economic growth rate and the U.S.current account balance are used as control variables at the same time or when the control variables are added again,the correlation between the U.S.dollar index and the U.S.federal bond yield will decrease significantly,indicating that when the U.S.federal bond yield rises,the decline in the international status of the dollar is weakened,and vice versa.When the control variables are increased from two or less to three or more control variables,the Copula function of the US dollar index and the US federal bond yield changes from G90 to C90,and the function changes,but both show an asymmetric structure.At the same time,from the tree diagram of the unrestricted nodes,it is found that the proportion of the dollar in the foreign exchange reserves is an important connection point between the dollar index and the US federal bond yield,and it is an important factor affecting the dollar index.The US current account balance is also an important factor affecting the dollar index.The US economic growth rate is an important factor affecting the US federal bond yield.In this paper,the combination model is used to select the US dollar trend prediction model.After a simple feature engineering,a new column of variables is generated as the difference summary.In the five models of Br model,XGBoost model,Ridge model,Knn model and Rf model,through 3-fold cross-validation method and based on the four evaluation index values of MAE,RMSE,MAPE and SMAPE,after comprehensive comparison,the best total model selected is Br model.For each independent variable,the optimal sub-model is also selected according to the above method and combined with random adjustment and synthesis.Among them,the best sub-models for US federal bond yield,US fiscal deficit ratio,US GDP growth rate,US current account balance,US capital account balance,US dollar ’s proportion in foreign exchange reserves and balance summary are Br model,Prophet model,Rf model,ARIMA model,Prophet model,ARIMA model and Rf model respectively.According to the above best sub-model,the first three quarterly values of each variable in 2022 are calculated,and these values are substituted into the Br model.The comprehensive values of the US dollar index in the first three quarters of 2022 are 1.7903,3.2562 and 4.7315 respectively.Through comparison,it is found that the calculated value is closer to the real value than the predicted value of the single smoothing method,indicating that the Br model is the optimal model to predict the trend of the dollar.