基于混合模糊隶属度的模糊双支持向量机研究

摘要

双支持向量机是近年提出的一种新的支持向量机.在处理模式分类问题时,双支持向量机速度远远超过传统支持向量机,而且显示出较好的推广能力.但双支持向量机没有考虑不同输入样本点可能会对分类超平面的形成产生不同影响,在某些实际问题中具有局限性.为了克服这个缺点,提出了一种基于混合模糊隶属度的模糊双支持向量机.该算法设计了一种结合距离和紧密度的模糊隶属度函数,给不同的训练样本赋予不同的模糊隶属度,构建两个最优非平行超平面,最终实现二值分类.实验证明,该模糊双支持向量机的分类性能优于传统的双支持向量机.%As a new version of support vector machine(SVM) ,twin support vector machine( TWSVM) is proposed recently. TWSVM is not only more faster than a conventional SVM, but shows good generalization for pattern classification. But the different effects of the different training samples on the classification hyperplanes are ignored in TWSVM,and the limitation is existed for some actual applications. Therefore, this paper presented a fuzzy twin support vector machine based on hybrid fuzzy membership. It designed a fuzzy membership function combined distance with affinity, and modified TWSVM by applying the fuzzy membership to every training sample. Finally it built two optimal nonparallel hyperplanes to achieve classification. The experiments indicate that the classification performance of the algorithm is more superiorer than a traditional TWSVM.