Twin proximal least squares support vector regression machine based on heteroscedastic Gaussian noise

Abstract

Twin proximal support vector regression is a new regression machine designed by using twin support vector machine and proximal support vector regression. In this paper, we use the above models framework to build a new regression model, called the twin proximal least squares support vector regression model based on heteroscedastic Gaussian noise (TPLSSVR-HGN). The least square method is introduced and the regularization terms b 1 2 and b 2 2 are added respectively. It transforms an inequality constraint problem into two simpler equality constraint problems, which not only improves the training speed and generalization ability, but also effectively improves the forecasting accuracy. In order to solve the parameter selection problem of model TPLSSVR-HGN, the particle swarm optimization algorithm with fast convergence speed and good robustness is selected to optimize its parameters. In order to verify the forecasting performance of TPLSSVR-HGN, it is compared with the classical regression models on the artificial data set, UCI data set and wind-speed data set. The experimental results show that TPLSSVR-HGN has better forecasting effect than the classical regression models.