TWSVR: Regression via Twin Support Vector Machine

Taking motivation from Twin Support Vector Machine (TWSVM) formulation, Peng (2010) attempted to propose Twin Support Vector Regression (TSVR) where the regressor is obtained via solving a pair of quadratic programming problems (QPPs). In this paper we argue that TSVR formulation is not in the true spirit of TWSVM. Further, taking motivation from Bi and Bennett (2003), we propose an alternative approach to find a formulation for Twin Support Vector Regression (TWSVR) which is in the true spirit of TWSVM. We show that our proposed TWSVR can be derived from TWSVM for an appropriately constructed classification problem. To check the efficacy of our proposed TWSVR we compare its performance with TSVR and classical Support Vector Regression(SVR) on various regression datasets.

[1]  Reshma Khemchandani,et al.  Twin Support Vector Machines for Pattern Classification , 2007, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[2]  Vladimir Cherkassky,et al.  Learning from Data: Concepts, Theory, and Methods , 1998 .

[3]  Alexander Gammerman,et al.  Ridge Regression Learning Algorithm in Dual Variables , 1998, ICML.

[4]  Yuan-Hai Shao,et al.  Improvements on Twin Support Vector Machines , 2011, IEEE Transactions on Neural Networks.

[5]  Olvi L. Mangasarian,et al.  Nonlinear Programming , 1969 .

[6]  Jinbo Bi,et al.  A geometric approach to support vector regression , 2003, Neurocomputing.

[7]  Corinna Cortes,et al.  Support-Vector Networks , 1995, Machine Learning.

[8]  Xinjun Peng,et al.  TSVR: An efficient Twin Support Vector Machine for regression , 2010, Neural Networks.