K-nearest neighbor based structural twin support vector machine

Structural twin support vector machine (S-TSVM) performs better than TSVM, since it incorporates the structural information of the corresponding class into the model. However, the redundant inactive constraints corresponding to non-support vectors (non-SVs) are still the burden of the solving process. Motivated by the KNN trick presented in the weighted twin support vector machines with local information (WLTSVM), we propose a novel K-nearest neighbor based structural twin support vector machine (KNN-STSVM). By applying the intra-class KNN method, different weights are given to the samples in one class to strengthen the structural information. For the other class, the superfluous constraints are deleted by the inter-class KNN method to speed up the training process. For large scale problems, a fast clipDCD algorithm is further introduced for acceleration. Comprehensive experimental results on twenty-two datasets demonstrate the efficiency of our proposed KNN-STSVM.