A preprocessing method for accelerating support vector regression training is presented in this paper. Then we propose a method of support vector regression by modifying the regularized risk fijnction. A data set from Shanghai Stock Exchange is used for the experiments to test the validity of our methods