A Novel Semismooth Newton Algorithm for PCA-Twin Support Vector Machine

Abstract

TWSVM is an algorithm model with good classification performance in machine learning. By finding two non-parallel hyperplanes, the model makes any hyperplane as close as possible to one kind of data and as far as possible to the other kind of data. In order to alleviate the influence of quadratic coefficient on the operation time during iteration, PCA dimensionality reduction algorithm is added in this paper. The semismooth newton algorithm is applied to solve dimensionality reduction L2-CSVM and L2-TWSVM. The results show that the proposed algorithm can not only significantly reduce the computational complexity, but also improve the training accuracy.