摘要
在ICA与聚类分析的基础上提出了一种改进的支持向量机分类模型——ICSVM模型。ICSVM模型中利用一种指标筛选算法与独立成分分析的方法将各数据指标转化为互相独立成分的数据指标。接着运用K-means方法对独立成分样本数据集进行聚类分析,再由获得的各子类中心数据构造初始的超平面,筛选出靠近初始超平面的支持类与亚支持类,并展开支持类与亚支持类中的样本数据点重新构造超平面,以便对数据进行分类。实验表明,对于样本比较多的数据集,与标准的SVM算法相比,ICSVM算法能够节约训练时间,同时能够提高分类的正确率。
Based on ICA and cluster analysis,this paper proposes ICSVM model.ICSVM model makes use of a selecting indices’ algorithm and ICA to transform the correlative indices into independent indices firstly.Then an initial classes is got by K-means cluster,and an initial super plane is made through the center of all subclass.By that the support classes and sub-support classes neighboring the initial super plane can be selected.Then expand the sample data of the support classes and sub-support classes and build up a new super plane by using them.Thus data can be classified by the new super plane.Compared with standard SVM,ICSVM has both better correct rate of classification and better training speed of ICSVM.
出处
《计算机工程与应用》
CSCD
北大核心
2008年第8期169-171,共3页
Computer Engineering and Applications
基金
广东省科技公关计划( the Key Technologies R&D Program of Guangdong Province, China under Grant No.B10101033, No.A10202001)
广东省哲学社会科学十一五规划项目( No.06M03)
华南农业大学校长基金( No.4900- K06166)
关键词
支持向量机
独立成分分析
聚类分析
相关关系
Support Vector Machine（SVM）
Independent Component Analysis（ICA）
cluster analysis
correlation
