SUPPORT VECTOR REGRESSION VIA MATHEMATICA

Abstract

In this tutorial type paper a Mathematica function for Support Vector Regression has been developed. Summarizing the main definitions and theorems of SVR, the detailed implementation steps of this function are presented and its application is illustrated by solving three 2D function approximation test problems, employing a stronger regularized universal Fourier and a wavelet kernel. In addition, a real world regression problem, forecasting of the peak of flood-wave is also solved. The %numeric and symbolic results show how easily and effectively Mathematica can be used for solving SVR problems.