一种基于DGMF与PSO的多核SVM  

A kind of Multiple Kernel Support Vector Machines Based on Double Gaussian Kernel Mixed-function and Particle Swarm Optimization

作　　者：田冬阳[1] TIAN Dong-yang（Wulumuqi Central Sub-branch of the People＇s Bank of China,Wulumuqi 830002,China）

机构地区：[1]中国人民银行乌鲁木齐中心支行,新疆乌鲁木齐830002

出　　处：《电脑知识与技术》2012年第5期3171-3178,共8页Computer Knowledge and Technology

摘　　要：目前常用的支持向量机采用传统的单一高斯核函数,或者传统高斯核函数与多项式核函数的组合,但是存在着分类器泛化推广能力不强,容易出现数据偏斜等问题。鉴于此,提出了基于DGMF的多核支持向量机、并采用粒子群算法优化其参数。最后构造了文本倾向性分类实验,实验结果表明改进算法在正确率、准确率、召回率、F1值上具有更好的性能。At present,the majority of text sentiment classification algorithms based on support vector machines using a single traditional Gaussian kernel function,or a combination of Gaussian kernel function and polynomial kernel function.But there are some problems： gen eralization abilities of these classifiers are not strong,the effects of fine-tune are not obvious.In view of this,experimental results show that the algorithm has better performances based on multiple kernel support vector machines of double Gaussian kernel mixed-function and par ticle swarm optimization algorithm.

关 键 词：双高斯合成函数 多核支持向量机 组合核函数 粒子群算法 文本倾向性分类 

分 类 号：TP181[自动化与计算机技术—控制科学与工程;自动化与计算机技术—控制理论与控制工程]