Nonparallel Quadratic Surfaces Support Vector Machine for Binary Classification

This paper introduces a novel kernel-free nonparallel quadratic surfaces support vector machine (NPQSSVM) designed for binary classification problems. Compared to existing quadratic surfaces and nonparallel classifiers, our proposed model offers several significant advantages: (a) it eliminates the need for selecting kernel functions and their associated parameters in addressing nonlinear classification tasks; (b) by employing an enhanced alternating direction method of multipliers, the NPQSSVM ensures scalability to large-scale classification problems involving both numerous instances and features; (c) it demonstrates stronger sparsity properties compared to typical quadratic surfaces support vector machine. Experimental results on a wide range of datasets consistently demonstrate that the NPQSSVM outperforms other algorithms in terms of efficiency and accuracy.