Application of SVM Based on Hybrid Kernel Function in Heart Disease Diagnoses
Yuanbin Mo, Shuihua Xu · 2010
Aimed at heart disease diagnose is an important issue and hybrid kernel functions have excellent learning ability and generalization performance, we propose SVM based on hybrid kernel function and apply the model to test the heart disease dataset. In this paper, K-type kernel function combine with linear kernel and polynomial kernel is firstly proposed, Linear combinations with different kernel functions are constructed and PSO algorithm is used to optimize the penalty parameter C. At last, the comparison of SVM with the kernel of this paper with the SVM with general kernel is given, and the results show that the SVM with the kernel of this paper has better performance.