带有惩罚项的多项式样条函数利率期限结构模型实证比较  

Empirical comparison term structure of interest rates model based on polynomial spline function with penalty term

作　　者：杨丰梅[1] 任姝仪[1] 周荣喜[2] 

机构地区：[1]北京化工大学理学院,北京100029 [2]北京化工大学经济管理学院,北京100029

出　　处：《系统工程理论与实践》2011年第4期735-739,共5页Systems Engineering-Theory & Practice

基　　金：国家自然科学基金(70701003;70801003);中央高校基本科研业务费专项资金(ZZ0915;ZZ1017);北京化工大学学科建设项目

摘　　要：针对多项式样条函数利率期限结构模型在曲线近端存在过度拟合的问题,首先提出了带惩罚项的自适应半参数回归方法来确定拟合函数的未知参数,同时应用广义交叉验证法确定正则化参数,并利用遗传算法求解惩罚因子的最优值.其次与多项式样条函数利率期限结构模型进行了实证比较,结果表明:所给模型能够提高曲线拟合的平滑度,但可能降低曲线的拟合精度.For the problem of overfitting the proximal curve in the term structure of interest rates model of polynomial spline functions,the adaptively semi-parametric regression with a penal term is developed to estimate the unknown parameters.The generalized cross-validation method is used to choose the smoothing parameter,and genetic algorithm is applied to look for the optimal smoothing parameter.Finally,the empirical comparison is made with the term structure of interest rates model based on polynomial spline functions.The results show that the model in this paper can improve the curve fitting smoothness,but possibly decrease the curve fitting accuracy.

关 键 词：利率期限结构 惩罚函数 GCV方法 遗传算法 

分 类 号：F830[经济管理—金融学]