Total Least Square Support Vector Machine for Regression

In this paper, we proposed a new support vector machine for linear regression in the total least square sense (TLSSVR). Mathematical technique taken in the total least square situations is more complicated than traditional least square method. Our new algorithm deriving from TLS-SVR results in solving a nonlinear equations by using Newton method. The results of computer simulations are given to illustrate that this TLS-SVR outperforms the commonly used least square regression.