Title :
Twin Support Vector Machine based Regression
Author :
Khemchandani, Reshma ; Goyal, Keshav ; Chandra, Suresh
Author_Institution :
Dept. of Comput. Sci., South Asian Univ., New Delhi, India
Abstract :
Taking motivation from Twin Support Vector Machine (TWSVM), Peng (2009) attempted to propose Twin Support Vector Regression (TSVR) where regressor was obtained via solving pair of Quadratic Programming Problems(QPPs). However the discussed formulation was not on the lines of TWSVM and had some restrictions. In this paper we propose formulation termed as Twin Support Vector Machine based Regression(TWSVR). Working on the lines of Bi and Bennett (2003), we derive this formulation from its classification counterpart TWSVM, i.e we have shown that TWSVR can be regarded as a classification problem, solution of whose is obtained by solving TWSVM. To check the efficacy of TWSVR we have compared its performance with TSVR and standard Support Vector Regression on various regression datasets.
Keywords :
quadratic programming; regression analysis; support vector machines; QPP; TWSVR; quadratic programming problems; twin support vector machine based regression; Kernel; Linear programming; Standards; Support vector machines; Testing; Training; Vectors; Machine Learning; Regression; Support Vector Machines; Twin Support Vector Machines;
Conference_Titel :
Advances in Pattern Recognition (ICAPR), 2015 Eighth International Conference on
Conference_Location :
Kolkata
DOI :
10.1109/ICAPR.2015.7050651