基于总间隔的模糊v-相对间隔机的研究  

Research of Fuzzy v-Relative Margin Machine Based on Total Margin

作　　者：龙章勇[1] 刘建华[2] 卢涵宇[3,4] 

机构地区：[1]南京铁道职业技术学院通信信号学院,江苏南京210015 [2]浙江工商职业技术学院工学院,浙江宁波310053 [3]贵州大学计算机与信息工程系,贵州贵阳550003 [4]成都理工大学信息科学与技术学院,四川成都610059

出　　处：《微电子学与计算机》2012年第6期167-171,共5页Microelectronics & Computer

基　　金：国家"八六三"计划项目(2006AA06Z110);国家科技支撑计划项目(2009BAH50B02)

摘　　要：类间间隔和类内聚类性是影响分类器分类性能的两种重要因素.基于模糊支持向量机和总间隔思想,提出一种基于总间隔的模糊v-相对间隔机(TMF-vRMM),本方法本质上是传统相对间隔机(RMM)的扩展,但可取得比RMM更好的分类性能.TMF-vRMM通过使用差异成本和引入总间隔和模糊隶属度,同时解决了不平衡训练样本问题和传统软间隔分类机RMM的过拟合问题,显著提升学习机的泛化能力.分别采用人造和实际数据集进行分类实验,结果显示TMF-vRMM具有优于相关方法的稳定分类性能.The between-class margin and within-class cohesion are two important factors impacting on the performance of classifiers. This paper presents a novel classifier called total margin based fuzzy v-relative margin machine （TMF-vRMM） based on the idea of fuzzy support vector machine （FSVM） and total margin. Although it can be seen as a modified class of the classic RMM, TMF-vRMM has better theoretical classification performance than RMM. The proposed method solves not only the over-fitting problem resulted from outliers with the approaches of fuzzification of the penalty and total margin algorithm, but also the imbalanced datasets by using different cost algorithm, thus obtaining a lower generalization error bound. Experimental results obtained with synthetic and real datasets respectively show that the algorithm proposed in the paper is stable and superior to other related diagrams.

关 键 词：相对间隔 总间隔 模糊 支持向量机 核方法 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]