岭回归在Nelson-Siegel利率期限结构中的应用

发布时间：2018-01-15 11:33

本文关键词：岭回归在Nelson-Siegel利率期限结构中的应用　出处：《西南财经大学》2013年硕士论文　论文类型：学位论文

【摘要】：利率期限结构一直是金融研究领域的一个重要方向。实践中Nelson-Siegel模型是最常用的利率期限结构模型。学界常用网格搜索法和普通最小二乘法来估计参数。然而,由于解释变量之间存在很强的多重共线性,这两种估计方法非常不稳定,得到一些无法合理解释的估计值,且方差较大。岭回归分析是一种专用于共线性数据分析的回归方法,所以本文选用岭回归来解决这些问题。上交所国债交易活跃,对市场信息敏感,适合利率期限结构的拟合。本文选取了2009年到2012年期间共968天的上交所国债交易数据为样本。论文中比较了网格搜索法、普通最小二乘法和岭回归估计法得到的参数序列。样本内拟合结果显示网格搜索法波动极大,且出现许多违背经济学理论的奇异值,而普通最小二乘法波动也较大。样本外预测结果表明岭回归在远端和近端都得到了最小平均绝对百分误差。本文的创新在于首先把岭回归引入Nelson-Siegel模型,为上海证券交易所国债利率期限结构拟合提供更有效的方法。
[Abstract]:Term structure of interest rate has always been an important direction in the field of financial research. In practice, Nelson-Siegel model is the most commonly used term structure model of interest rate. To estimate the parameters by multiplying. Because there is a strong multiple collinearity between the explanatory variables, these two methods are very unstable, and some estimates can not be reasonably explained. Ridge regression analysis is a kind of regression method for collinear data analysis, so this paper chooses Ridge regression to solve these problems. Shanghai Stock Exchange treasury bonds trading is active, sensitive to market information. This paper selects 968 days of SSE treasury bond trading data from 2009 to 2012 as a sample. The paper compares the grid search method. The fitting results in the sample show that the grid search method fluctuates greatly and many singular values which are contrary to the economic theory appear. The prediction results show that the minimum mean absolute percent error of ridge regression is obtained at the far end and the proximal end. The innovation of this paper is to introduce ridge regression into Nelson-S first. Iegel model. It provides a more effective method for fitting the term structure of treasury bond interest rate in Shanghai Stock Exchange.
【学位授予单位】：西南财经大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F224;F832.5

【参考文献】