摘要
针对面向分类去噪问题,提出了一种新的模糊支持向量机算法(ν-FSVM),并给出了通过无穷次连续可微函数建立模糊关系的方法.该方法能对训练集中的点自动赋予模糊关系,并且对带有噪声的点和孤立的点赋予较小的模糊关系.与传统的ν支持向量机比较,该算法通过建立训练集的模糊关系,能够大大减小噪声对分类的影响,从而提高分类精度,减少误差.
A fuzzy support vector machine learning approach, namely ν-FSVM is presented to eliminate classification noise. And the fuzzy relations among the training data points are also created with an infinite continue differentiable function, which enables to automatically assign the fuzzy relationships to the training data points, and assign small fuzzy relationships to the points with noises or outliers. Comparing with the traditional ν-support vector machine, the experimental results on benchmark datasets show that the presented approach reduces the outlier effects to improve the classification performance.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2007年第12期1414-1417,共4页
Journal of Xi'an Jiaotong University
基金
国家自然科学基金资助项目(10671153)
关键词
支持向量机
分类
ν支持向量机
support vector machine
classification
ν-support vector machine
