Research On Forecasting And Trading Strategy Of Treasury Yield Curve Using Functional Data Analysis

Since the issuance of treasury bonds in 1981,the treasury bond trading market has developed rapidly.As a kind of investment tool for fixed-income markets,treasury bonds have been favored by more and more investors.The bond yield curve is used as a curve to describe the relationship between bond yield to maturity and its remaining maturity.Accurate estimation is important for researchers to further study the term structure of interest rates and help investors make bond investment decisions.However,traditional data analysis methods can no longer satisfy people's research on massive high-frequency financial data,and functional data analysis methods,as a new data analysis method,have certain advantages in dealing with massive high-frequency data.Therefore,it is of great significance to use the functional data analysis method to forecast the yield curve.This paper uses functional data analysis to forecast the yield curve of China.To reflect the advantages of the model,we also use the traditional NSS-ARMA model to forcast the yield curve.By comparing the sum of squared residuals of two models,we can conclude that functional data analysis is more suitable for forecasting the yield curve.Based on the forecast result,we propose a new bond trading strategy,in order to provide investors with reference for treasury bonds trading.The main conclusions of this paper are as follows:(1)Using the functional autoregressive model to forecast the yield curve of treasury bonds,the forecasting accuracy is higher than the traditional NSS-ARMA model,and the forecast result basically fits the actual curve,and The accuracy of the forecast does not decline as the forecast period elapses.(2)The forecast-based bond trading strategy is significantly better than the benchmark index in terms of investment income.(3)Forecast-based bond trading strategies are highly real-time.Since the functional autoregressive model can be used to realize the real-time forecast of the daily bond yield curve,the forecast-based bond trading strategy can select different trading strategies according to different daily real-time forecasting yield curves,and further select the daily trading strategy.Trading on the most undervalued bond has a stronger reference for investors to make investment decisions.The innovation of this paper is that on the one hand,the functional data analysis method is applied to the forecast analysis of the yield curve,which makes up for the parameter estimation that the traditional model needs to repeat for each period,and the accuracy of the forecast with the accuracy of the forecast period.On the other hand,the traditional trading strategy only provides investors with a range of tradable bonds,and this paper proposes a method for selecting specific trading vouchers based on the forecast.