Abstract

The present study suggest modified twin support vector regression (MTSVR) for data regression. In the MTSVR model, the regression function is determined using a pair of unparalleled up and down bound functions. In any optimization problem, a new term is added to obtain structural information of the input data based on the concept of structural granularity. Furthermore, Successive Over relaxation is used to accelerate the training process of optimization problems. Particle Swarm Optimization (PSO) algorithm is used to determine the parameters of the MTSVR model. According to the results of the artificial and real datasets, the prediction accuracy and generalization capability of the MTSVR model is significantly increased.