A ν-twin support vector machines based on minimum class variance

Twin support vector machine (TWSVM) is an effective tool for data classification and regression since it solves a pair of smaller-sized quadratic programming problems (QPPs) rather than a single large one in a classical support vector machine (SVM), and thus TWSVM derives better training speed than a classical SVM. In this paper, we propose a new version of twin support vector machine, a minimum class variance ν-twin support vector machine (ν-MCVTWSVM), to improve upon the TWSVM. This ν-MCVTWSVM introduces a pair of parameters (ν) to control the bounds of the fractions of the support vectors and the error margins. Numerical experiments on some real-world databases demonstrate that the superiority of the proposed method over the classical learning paradigm in terms of both classification accuracy and the number of support vectors. © 2013 by CESER Publications.