论文部分内容阅读
针对C-支持向量机(C-SVM,C-Support Vector Machine)中惩罚系数C可能导致最优分类面不合理的问题,提出基于误差最小的SVM最优分类面修正方法.通过调整正负类分类间隔的约束条件,求解使训练样本总误差最小的偏置系数,并兼顾与正负类误差之差的绝对值的平衡,得到误差最小的更优分类面.实验证明该修正方法与C-SVM及其它修正方法相比,具有较高的分类精度和较强的抗噪声与野值数据干扰能力.
Aiming at the problem that the penalty coefficient C in C-Support Vector Machine (C-Support Vector Machine) may lead to the irrational classification, a new SVM method based on the smallest error is proposed. By adjusting the positive and negative classes And the classification interval, the bias coefficient that minimizes the total error of the training samples and the balance of the absolute value of the difference between the positive and negative errors are obtained, and the better classification surface with the smallest error is obtained.The experiment proves that the correction method is similar to the C- Compared with other correction methods, SVM has higher classification accuracy and strong anti-noise and outliers data interference ability.