中文摘要：

本文从经典利率期限结构模型和数量利率期限结构模型中选取了三因子Vasicek模型、三因子CIR模型、多项式样条模型、指数样条模型、DL模型和动态SV模型等6个应用广泛且具有代表性的利率期限结构模型,分别基于2008年7月至2014年3月中国和美国市场的月度国债收益率数据进行拟合与预测,并采用均方误差（RMSE）和平均绝对误差（MAE）对实证效果进行判别.结果表明,DL模型在针对中国市场和美国市场数据的拟合与预测方面能力均十分突出且效果稳定,指数样条模型次之,而其他模型则在利率期限结构特征的刻画效果方面存在更强的数据依赖与能力不足问题.本文的结论能够为实证研究利率期限结构特征的模型选取提供科学依据和数据支持.

英文摘要：

This paper selects six common and representative models from the classical and quantitative models of the term structure of interest rates, including three-factor Vasicek model, three-factor CIR model, polynomial spline model, exponential spline model, DL model and dynamic SV model, then makes a fitting and forecasting based on the China and America monthly treasury yields from July 2008 to March 2014, and compares the empirical capacity using the mean square error and mean absolute error. The results show that, DL model＇ s fitting ability and prediction ability are both prominent and stable no matter for the China market or America market, and the exponential spline model follows, other models show stronger data-dependent and greater lack of capacity problems. The conclusion can provide scientific basis and data support for model selecting when do the research of the term structure of interest rates.