Efficient Cross Validation for SVR Based on Center Distance in Kernel Space

Abstract

Cross validation (CV) is widely used to find the optimal parameters of the support vector regression (SVR) model. Regarding the conventional CV method, the optimal model parameters may be affected when the training set is randomly split into k disjoint folds. In the paper, an efficient CV based on center distance in kernel space is presented. Data splitting is based on the distance between the sample and the center point in the kernel space. Simulation experiments results show that the proposed CV method makes the selection of optimal model parameters more reasonable and improves the generalization ability of SVR model.