基于邻域分布局部线性嵌入的粗集支持向量机  

ROUGH SET SUPPORT VECTOR MACHINE BASED ON NEIGHBOURHOOD DISTRIBUTION LOCALLY LINEAR EMBEDDING

作　　者：刘立群[1] 韩俊英[1] 刘成忠[1] 

机构地区：[1]甘肃农业大学信息科学技术学院,甘肃兰州730070

出　　处：《计算机应用与软件》2014年第6期264-266,284,共4页Computer Applications and Software

基　　金：国家自然科学基金项目(61063028);甘肃省自然科学研究基金计划项目(1208RJZA133);甘肃省高等学校研究生导师科研项目(1202-04);甘肃省青年科技基金计划项目(1208RJYA096);甘肃省科技支撑计划项目(1011NKCA058)

摘　　要：针对粗集支持向量机面对高维特征值的训练样本分类效率低的问题,采用邻域点间的分布矩阵用于改进局部线性嵌入的权值矩阵重建,在粗集支持向量机上引入该邻域分布局部线性嵌入算法,对输入训练集进行数据降维操作。对三个数据集分别采用传统支持向量机、粗集支持向量机和基于邻域分布局部线性嵌入的粗集支持向量机进行数据分类测试,结果表明改进的支持向量机对于高维特征值的训练样本,可有效提高分类准确率及计算效率。To solve the problem of low classification efficiency the rough set support vector machine （ RSSVM ） has when encountering the training samples with high-dimensional eigenvalue, we adopt the distribution matrix between the neighbourhood points to improve the recon- struction of LLE weight matrix. The neighbourhood distribution locally liner embedding （NDLLE） algorithm is introduced into RSSVM to re- duce the dimensionality of input training data set. In experiments, three data sets are tested their data classification with traditional SVM, RSSVM and NDLLE-RSSVM respectively. The results indicate that the NDLLE-RSSVM can effectively improve classification accuracy and computing efficiency for the training sample with high-dimensional eigenvalue.

关 键 词：粗集支持向量机 局部线性嵌入 邻域点 分布矩阵 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]