Author(s)

Hua Zheng, Dongzhu Zhao, Yafei Shang and Shiqiang Duan

Corresponding Author

Hua Zheng

Abstract

Based on the structural risk minimization criterion, an online Support Vector Machine (SVM) algorithm with adaptively selecting kernel functions is presented. In order to overcome the problem that the traditional method converges slowly and cannot adaptively select samples, this paper first sets the appropriate window function for the time series signal, and then selects the appropriate truncation error minimization criterion of the Least Square Support Vector Machine (LS-SVM) in the process of sample update. The Lagrange factor finally completes the retraining of the new samples. Compared with the traditional SVM method, the results of numerical simulation show that, the proposed algorithm has the characteristics of high prediction accuracy and strong generalization ability, and could be widely used in a series of engineering applications such as pattern recognition, fault diagnosis, machine vision and intelligent control.

Keywords

Adaptive Kernel Function; Truncation error minimization criterion; Structural risk minimization; Least Square Support Vector Machine