A method determining parameters of SVR model based on Probability and Statistics

To get over the difficulties in adopting conventional leaving-one cross validation method to decide the parameters of support vector regression (SVR) forecasting model for small sample with noise or errors, an advanced method was developed based on the theories of probability and statistics. The proposed method considered that the true parameters of the model should have much higher probability to gain better forecasting results than that of others. Living examples show that the SVR forecasting model with the parameters calculated by the method presented here has better results than that of other parameters