摘要
为了降低支持向量机对不平衡数据的倾向性影响,以及减弱其对噪声点或野值点的敏感,提出了一种新的模糊支持向量机隶属度函数设计方法.该方法分析产生倾向性的原因,有效地区分样本对分类面的贡献,合理地设计隶属度函数.最后通过对含噪声的非均衡数据实验表明,该方法平衡了倾向性,提高了预测分类精度,从而增强了支持向量机在入侵检测和故障诊断等方面的应用.
To eliminate the disadvantage of the traditional SVM that the classifier of unbalanced data is unfair to the rare class,a new fuzzy membership function of fuzzy support vector machine is presented. The proposed method analyses the reasons of the unfair theoretically and weights the contribution of samples to the separating plane, designs the membership function reasonably. Numerical experiments for unbalanced data sets with noise show that the proposed algorithm balances the classifier and achieves the better classification accuracy than SVM. It boosts up the performance of the application of SVM in network intrusion detection and fault detection.
出处
《西安工业大学学报》
CAS
2008年第3期297-300,共4页
Journal of Xi’an Technological University
基金
国家自然科学基金项目(60574075)
关键词
支持向量机
模糊支持向量机
非平衡数据
分类精度
隶属度函数
support vector machine,fuzzy support vector machine,unbalanced data
classification accuracy
membership function
