样条函数在构造债券收益率曲线中的应用

摘要

针对债券市场上芜杂的行情数据，提出将DBSCAN聚类算法应用于构造债券收益率曲线样条函数。通过运用DBSCAN算法对用于构造债券收益率曲线的行情数据进行聚类分析，能够有效地剔除市场上的异常交易数据。在聚类分析结果的基础上，再次应用DBSCAN算法于构造债券收益率曲线，根据市场上行情数据的密集区域对样条函数进行分段。此外，针对传统的依赖于经验进行债券收益率曲线样条函数分段点选取的缺点，使用DBSCAN算法可有效地提高债券收益率曲线和行情数据的符合程度。实验结果表明，将DBSCAN算法用于构建债券收益率曲线样条函数，可以提高收益率曲线反映利率期限结构波动及准确性的效果。%In light of the miscellaneous quotation data in bond market,we proposed to apply the DBSCAN algorithm to constructing the spline function of bond yield curve.By employing DBSCAN algorithm to the cluster analysis of the quotation data,which is used to construct bond yield curve,it is able to eliminate effectively the abnormal transaction data on the market.On the basis of cluster analysis results,we used the DBSCAN algorithm once again to construct the bond yield curve,and segmented the spline function into sections according to the dense regions of quotation data on the market.Besides,targeted at the defect of traditional way that it selects the segment points of spline function of bond yield curve depending on experience,the use of DBSCAN algorithm can effectively improve the conformance between the bond yield curve and the bond quotation data.From the experimental results it is illustrated that to apply the DBSCAN algorithm to constructing the spline function of bond yield curve can improve the effects of reflecting the fluctuations of interest term structure by bond yield curve and its accuracy.