基于总间隔与代价差异算法的支持向量机

Research on SVM Based on Total Margin and Different Cost Algorithm

针对支持向量机在实际应用中存在的最优分类超平面的倾斜问题和推广误差界的问题,引入了总间隔与代价差异算法,对标准的支持向量机算法进行了改进.同时,针对线性和非线性两种情况,给出了详细的公式推导过程,并得出结论:基于总间隔与代价差异算法的支持向量机的性能优于标准的支持向量机.

In respect of the skew of the optimal separating hyperplane and the generalization error bound that exist in the practical application of support vector machine, this paper introduces the total margin algorithm and different cost algorithm, which improves the standard support vector machine. Meanwhile, a detailed derivation of the formula in both linear and nonlinear cases is showed in the paper. And a conclusion can be made that the support vector machine based on total margin and different cost algorithm provides good performance, comparing with the standard support vector machine.