Improved Nonparallel Hyperplanes Support Vector Machines for Multi-class Classification

In this paper, we present an improved nonparallel hyperplanes classifier fbr multi-class classification, termed as IN-HCMC. As in the nonparallel support vector machine (NPSVM) for binary classification, the epsilon-insensitive loss function is adopted in the primal problems of multi-class classification to improve the sparseness associated with the nonparallel hyperplanes classifier for multi-class classification (NHCMC) where the quadratic loss function is used. Experimental results on some benchmark datasets are reported to show the effectiveness of our method in terms of sparseness and classification accuracy.