Robust and Sparse Twin Support Vector Regression via Linear Programming
Xiaobo Chen, Jian Jun Yang, Jun Sheng Liang, Qiaolin Ye · 2010
Twin support vector regression (TSVR) was proposed recently as a novel regressor that tries to find a pair of nonparallel planes, i.e. ε-insensitive up- and down-bound, by solving two related SVM-type problems. Though TSVR exhibits good performance compared with conventional methods like SVR, it suffers from several issues. In this paper, we propose a novel regression algorithm called Robust and Sparse Twin Support Vector Regression (RSTSVR). The idea is to reformulate TSVR as a strongly convex problem by regularization technique firstly and then derive a linear programming (LP) formulation which is not only simple but also introduces robustness and sparseness. Instead of solving the resulting LP problem straightforward, we convert the primal LP to its dual to simplify computation. The experimental results on several publicly available benchmark data sets show the feasibility and effectiveness of the proposed method.