Research on Combination Kernel Function of Support Vector Machine

• Published in: 2008 International Conference on Computer Science and Software Engineering Vol. 1; pp. 838 - 841
• Main Authors: Huazhu Song, Zichun Ding, Cuicui Guo, Zhe Li, Hongxia Xia
• Format: Conference Proceeding
• Language: English; Japanese
• Published: IEEE 2008
• Subjects: combination kernel function; Computer science; Gaussian radial basis kernel function (RBF); Kernel; linear kernel function; Machine learning; Neural networks; Pattern analysis; Polynomials; Support Vector Machine (SVM); Support vector machine classification; Support vector machines; Time series analysis; Virtual colonoscopy
• ISBN: 0769533361; 9780769533360
• DOI: 10.1109/CSSE.2008.1231

The kernel function and parameters selection is a key problem in the research of support vector machine. After discussing the influence of support vector machine on kernel parameters and error penalty factors, a new kernel function CombKer was proposed and constructed. The CombKer kernel function is a kind of combination kernel function, which combines the Gaussian RBF kernel function that has the local characteristic, with the linear kernel function that has the global characteristic. Finally, some experiments on different domains data in the support vector machine constructed by the CombKer kernel function were done, and the results showed the better ability on prediction of this kind of support vector machine and proved the validation of the CombKer kernel function.