SVM核函数对分类精度影响的研究

Impact of SVM Kernel Function on the Classification Accuracy

支持向量机是一种基于统计学习理论的新型机器学习算法,它通过求解最优化问题,在高维空间中寻找最优分类超平面,从而解决复杂数据的分类、回归问题.文中介绍了支持向量机的基本算法原理及其分类方法,重点研究将核函数引入不可分的情形.本文通过改变核函数的参数,采用对比实验来比较分类精度,同时根据Mercer条件形成新的线性组合核函数,最后得出通过改变核函数参数与线性组合核函数的方法可以明显提高分类的精度.

Support vector machine is a new kind of machine learning algorithm based on statistical learn- ing theory. It＇s started with searching the optimum solution in the high dimension space in search for the optimal classification hyperplanes, so as to solve complex data classification or regression problems. This paper intro- duced the basic algorithm of support vector machine principle, classification method and key researches that can be introduced into the kernel function indivisible case by changing the parameters of the kernel function. The ac- curacy of classification was compared with experiment. At the same time, according to Mercer conditions, a new linear combination kernel function was proposed. It is proved that changing kernel function parameter and linear combination of kernel function can improve the accuracy of classification.