Application Of Ns Model Factor Selection Technology In Interest Rate Term Structure

The term structure of interest rates is the basis of interest rate risk management,financial asset pricing and financial product design.Its estimation is an important research direction of finance.A common model for estimating the term structure of interest rates in the industry is the NS model and its derived SV model.Central bank banks,monetary policy makers,and financial institutions in many countries use the NS model to estimate the term structure of interest rates.The NS model is a model described by three factors,but the study found that the NS model can only fit a simple shape interest rate curve.When the interest rate curve is more complicated,the fitting ability of the NS model is limited,and the term structure of the interest rate cannot be accurately estimated.The SV model adds a hump factor to the three-factor of the NS model,which fits the bimodal and double U-type interest rate curves.This paper attempts to propose a model to fit a more complex interest rate curve,making it more accurate to estimate various complex interest rate term structures.In order to make the NS model have stronger fitting ability and inspire the SV model to add a hump factor,this paper adds multiple hump factors to the NS model to construct the NS model with multiple hump factors(ie LASSO-NS model).However,considering the addition of multiple hump factors,the model may have over-fitting problems.This paper introduces LASSO technology to select the number of factors in the model.The LASSO technique obtains the parameter estimates of the model while selecting the number of factors.The LASSO technique adds the Ll norm of the regression coefficient of each hump factor as a penalty term,and adjusts the value of the penalty coefficient to select the factor.In the factor selection process,the number of factors of the optimal model is automatically selected along with the change of the penalty coefficient value.And give the model's parameter estimates,so that the model has a stronger ability to fit the complex shape term structure.This paper selects the data of 66 trading days on the last trading day of each month in the Shanghai Stock Exchange's government bond market from January 2013 to June 2018 as an analysis sample.This paper compares the LASSO-NS model with the NS model and the SV model from four aspects:cross-sectional data,intra-sample fitting,out-of-sample prediction and parameter stability.The empirical results show that the LASSO-NS model is in these four aspects.The results are superior to the NS model and the SV model.The LASSO-NS model can flexibly represent a complex interest rate term structure,which not only overcomes the shortcomings of the SV model parameter estimation value depending on the given initial value,but also reflects the multi-peak interest rate curve well,in the fitting accuracy and The parameter stability is also superior to the NS model and the SV model,and it can more accurately estimate the term structure of the interest rate of China's national debt.