Research of the Term Structure of Interest Rates Based on Improved J-LSSVR

摘要

In this paper, Least Squares Support Vector Regression (LSSVR) method is improved to have a sparse, and enhance the generalization ability of the method. In LSSVR model, an increase of three indexes set down by certain criteria to determine the selection and support vector. Then two models were selected to fit the sample data on the bond spot yield curve. We can derived from analysis and comparison revealed that the improved model fit residuals and LSSVR model results broadly consistent with their training time and prediction time of a relatively large Shortened. Thus, we can set through the addition of three indexes to increase the generalization ability and reduce the computing time when using support vector machine to fit the term structure of interest rates.