Abstract: The paper classif the term structure of interest rate models into four categories and summarizes the estimation methods for interest rate models.The paper uses interest rate data of China and USA to demonstrate that the weak efficiency of interest rate market,and that estimation methods and numerical optimization methods affect the estimation results.Empirical results indicate that the new estimation method,which takes advantage of all information in all market interest rates data,can get the more accurate parameters and eliminate arbitrage in interest rate market.It also indicates that genetic method is not very stable;nelder-mead method is sensitive to initial value of these parameters and,divided rectangles method can get the most robust estimation results.

Key words: term structure of interest rate, martingale difference, genetic search, divided rectangles GMM