分别基于零息票债券的连续复利到期收益率(简称零息收益率)、零息票债券的半年复利平价收益率(简称平价收益率)和瞬时远期利率,对一类轨道随时间非减的Lévy过程――从属过程进行了参数估计和模型之间的比较.对于转移函数已知的从属过程,用极大似然估计得到了参数的估计.对于转移函数未知的从属过程,可以利用鞍点逼近的方法得到转移函数的近似表达式,从而利用极大似然函数进行参数估计.Gamma过程对三类收益率曲线的描述上优于稳定从属过程和泊松随机积分过程.相比较零息收益率和平价收益率,稳定从属过程和泊松随机积分过程在对瞬时远期利率的描述上要吻合一点. Based on the continuous compound interest-earning yield (zero-yield rate) of zero-coupon bonds, the semi-annual compound interest-free par value (the rate of return) and the instantaneous forward rate of zero-coupon bonds, Subtractive Lévy process - the subordinate process makes the parameter estimation and the comparison between the models.For the subordinate process whose transfer function is known, the maximum likelihood estimation is used to get the parameter estimation.For the subordinate process whose transfer function is unknown, Saddle point approximation method to get the approximate expression of the transfer function, and thus use the maximum likelihood function for parameter estimation.Gamma process for the three types of yield curve description is better than the stable subordinate process and Poisson random integration process.Compared with zero interest income Rate and parity rate of return, the process of stable subordination and Poisson random integration process in the description of the forward and forward interest rates to be consistent.