Laplacian twin support vector machine with Pinball loss for semi-supervised classification

Semi-supervised learning utilizes labeled data and the geometric information in unlabeled to construct a model whereas supervised makes use of only label data. So, semi-supervised establishes more reasonable classifier. In recent years, Laplacian support vector machine (Lap-SVM) has received lot interest framework classification. To develop performance Lap-SVM, twin (Lap-TSVM) shown exceptional as an addition improve computational complexity. However, dealing with noise sensitivity instability for resampling due its hinge loss function is still challenge. this paper, we provide novel by combining pinball function, termed Lap-PTSVM, effectively handle aforementioned problems. As result, it improves better generalization ability Several experiments have been performed on artificial UCI datasets. The results show that proposed insensitivity comparable Lap-TSVM great performance. Furthermore, non-parametric statistical test are also conducted justify competitive Lap-PTSVM.