基于模糊核聚类粒化的粒度支持向量机

针对传统的粒度支持向量机（granular support vector machine， GSVM）将训练样本在原空间粒化后再映射到核空间，导致数据与原空间的分布不一致，从而降低GSVM的泛化能力的问题，本文提出了一种基于模糊核聚类粒化的粒度支持向量机学习算法（fuzzy kernel cluster granular support vector machine， FKC-GSVM）。FKC-GSVM通过利用模糊核聚类直接在核空间对数据进行粒的划分和支持向量粒的选取，在相同的核空间中进行支持向量粒的GSVM训练。在UCI数据集和NDC大数据上的实验表明:与其他几个算法相比，FKC-GSVM在更短的时间内获得了精度更高的解。

Granular support vector machine based on fuzzy kernel clustering granulation

For the traditional granular support vector machine (GSVM), the training samples are granulated in the original space and then mapped to the kernel space. However, this method will lead to the inconsistent distribution of the data between the original space and the kernel space, thereby reducing the generalization of GSVM. To solve this problem, a granular support vector machine based on fuzzy kernel cluster is proposed. Here, the training data are directly granulated, and support vector particles are selected in kernel space. The support vector particles are then trained in the same kernel space by the GSVM. Finally, experiments on UCI data sets and NDC big data sets show that FKC-GSVM achieves more accurate solutions in a shorter time than other algorithms.