核对齐多核模糊支持向量机

Kernel-target alignment multi-kernel fuzzy support vector machine

支持向量机(SVMs)是当前被广泛使用的机器学习技术,其通过最优分割超平面来提高分类器的泛化能力,在实际应用中表现优异。然而SVM也存在易受噪声影响,以及核函数选择等难题。针对以上问题,本文将基于核对齐的多核学习方法引入到模糊支持向量机(fuzzy support vector machine,FSVM)中,提出了模糊多核支持向量机模型(multiple kernel fuzzy support vector machine,MFSVM)。MFSVM通过模糊粗糙集方法计算每一样例隶属度;其次,利用核对齐的多核方法计算每一单核权重,并将组合核引入到模糊支持向量机中。该方法不仅提高了支持向量机的抗噪声能力,也有效避免了核选择难题。在UCI数据库上进行实验,结果表明本文所提方法具有较高的分类精度,验证了该方法的可行性与有效性。

Support vector machines(SVMs)are widely used machine learning techniques.They are used to construct an optimal hyper-plane and have an extraordinary generalization capability and good performance.However,SVMs are sensitive to noise,and it is difficult to select an appropriate kernel for SVMs.In this paper,we introduce kernel-target alignment-based multi-kernel learning method into fuzzy support vector machine(FSVM)and propose the kernel-target alignment-based multi-kernel fuzzy support vector machine(MFSVM).First,we assign the corresponding membership degree to each sample point by the fuzzy rough set method,and then calculate the kernel weight by the multi-kernel learning based on the kernel alignment.Then,the combined kernel is introduced into the fuzzy SVM.The proposed method not only improves the anti-noise ability of the SVM but also effectively avoids the problem of kernel selection.Experiments on nine datasets of the UCI database show that the proposed method has a high classification accuracy,which verifies its feasibility and effectiveness.