Weighted Least Squares Twin Support Vector Machines for Pattern Classification

In this paper we propose a weighted version of recently developed least squares twin support vector machine (LSTSVM) for pattern classification, in which different weights are put on the error variables in order to eliminate the impact of noise data and obtain the robust estimation. Here, we offer the formulations of the proposed weighted LSTSVM (WLSTSVM) in both linear and nonlinear cases. Comparative experiments have been made on UCI datasets for different kernels, and the experimental results show that the proposed algorithm has better performance in testing accuracy than LSTSVM, while the computational complexity is stable.