Efficient parameter selection for Support Vector Regression using orthogonal array

Support Vector Regression (SVR) is a nonlinear prediction method using kernel function and well known to have high accuracy in prediction. In addition, it has been widely applied to real-world problems. Although the accuracy of an effectively tuned SVR is high, its performance strongly depends on hyperparameters given from outside of the model. Therefore, the determination of the parameters is important when applying SVR to real-world problems. Although the optimum parameters are usually determined by an exhaustive grid search, using this method is not realistic when the sample size is considerably large in big data analysis, because the execution of SVR requires more computational time as the number of samples increases. In order to decrease the computational time required to determine the optimum parameters, we conduct a particular sampling based on an orthogonal array and propose an efficient method for parameter tuning for SVR. The proposed method can reduce the computational time to approximately one-twelfth of that taken by a grid research. We validate the accuracy of the proposed method by applying it to a wine quality prediction problem. The results of the proposed method are ranked second among all the combinations of parameters sampled using grid search. In addition, its performance is superior to that of a random method.