论文部分内容阅读
对最小化数据集在超平面法向离散程度进行了研究,提出一种改进的支持向量机,称为最小法向离散度模糊支持向量机(minimal normal scatter fuzzy support vector machine,MNS-FSVM)。MNSFSVM通过使优化目标在最大化几何间隔与最小化数据集在超平面法向离散程度之间达到一个平衡来减小离群点对超平面的影响。对二个基准数据集进行了实验验证,结果表明,MNS-SVM不仅能够提高分类的准确率,对离群点也有很好的鲁棒性。
The discretization degree of the minimized dataset in the hyperplane is studied. An improved support vector machine is proposed, called the minimal normal scatter fuzzy support vector machine (MNS-FSVM). MNSFSVM reduces the impact of outliers on the hyperplane by making the optimization goal a balance between maximizing the geometric separation and minimizing the dataset at the hyperplane normal dispersion. Experimental results on two benchmark datasets show that MNS-SVM can not only improve the classification accuracy but also robustness to outliers.