摘要
核函数是支持向量机的核心,不同的核函数将产生不同的分类效果。由于普通核函数各有其利弊,为了得到学习能力和泛化能力较强的核函数,根据核函数的基本性质,两个核函数之和仍然是核函数,将局部核函数和全局核函数线性组合构成新的核函数——混合核函数。该核函数吸取了局部核函数和全局核函数的优点。利用混合核函数进行流程企业供应链预测实验,仿真结果验证了该核函数的有效性和正确性。
Kernel function is the key issue of Support Vector Machines ( SVM), and different kernel functionscan produce different SVM. As the general kernel functions have their own advantages and disadvantages, to get a kernel function with stronger learning and generalization ability, a new mixture kernel function was proposed based on the fundamental of the kernel function, that the combination of the kernel functions still be a kernel function. The new kernel function had the desirable characteristics for SVM learning and generalization, and learned the advantages of global kernels and local kernels. The comparison results between the new kernel and other kernels in forecast of process supply chain experiment certify that the new kernel can achieve better performance than other kernels.
出处
《计算机应用》
CSCD
北大核心
2009年第B12期173-175,178,共4页
journal of Computer Applications
基金
国家自然科学基金资助项目(60461001)
广西自然科学基金资助项目(0832082
0991086)
国家民委科研项目基金资助项目(08GX01)
关键词
支持向量机
核函数
全局核
局部核
基于核的学习
Support Vector Machine (SVM)
kernel function
global kernel
local kernel
kernd-based learning