Function approximation based on Twin Support Vector Machines

A new function approximation algorithm based on twin support vector machines (TSVM) is presented in this paper. Support vector regression (SVR) has been shown to have good robust properties against noise in function approximation, however, the overfitting phenomena cannot be eliminated if the parameters used in SVR are improperly selected, and the selection of various parameters is not straightforward. In this paper, we use the properties of TSVM to solve this problem, that it will generate two nonparallel planes such that each plane is closer to one of the two classes and is as far as possible from the other. The experiments show good performances without additional computing time by using traditional test functions, data with noise and ambiguous training data.