Asymmetric ε-band fuzzy support vector regression based on data domain description

To solve over-fitting problems of standard support vector machine(SVM) for the noise, a new Asymmetric ε - band fuzzy support vector regression based on data domain description (ASVDD) is presented by analyzing the principle of support vector regression and the characteristics of the data domain in this paper. Using it to forecast time series of airport fuel consumption and the predicted results are compared with standard support vector machine´s. Research results show that the Asymmetric ε - band fuzzy support vector regression based on data domain description (ASVDD) has a higher prediction precision on 2-dimensional data set simulation and airport fuel consumption time series than standard support vector machine.