Nonparallel least square support vector machine for classification

Nonparallel support vector machine based on one optimization problem (NSVMOOP) aims at finding two nonparallel hyper-planes by maximizing the intersection angle of their normal vectors w (1) and w (2). As maximum intersection angle preserves both compactness and separation of data, NSVMOOP yields good forecasting accuracy. However, as it solves one large quadratic programming problem (QPP), it costs high running time. In order to improve its learning speed, a novel nonparallel least square support vector machine (NLSSVM) is proposed in this paper. NLSSVM solves a linear system of equations instead of solving one large QPP. As both intersection angle and least square version are applied on our NLSSVM, it performs better generalization performance than other algorithms. Experimental results on twenty benchmark datasets demonstrate its validity.