关键词: 核超球, 最大边界, 核函数, 模糊

Abstract: In order to improve the classification accuracy of multiclass, an algorithm called Large Margin and Fuzzy Kernel Hyper-Ball (LMFKHB) was proposed. First, the sample datasets were mapped into a high-dimensional feature space through a kernel function. Then, all decision functions were obtained using the proposed method. Meanwhile, a fuzzy membership function was introduced to solve the wrong classification issue for these samples in the dead zone, thus the flexibility was enhanced and the classification accuracy was improved. The experiments on the artificial and real data demonstrate the effectiveness of the method.

Key words: Kernel Hyper-Ball, large margin, kernel function, fuzzy