基于粗糙集的支持向量回归机混合算法  

Support vector regression hybrid algorithm based on rough set

作　　者：邓九英[1,2] 王钦若[1] 毛宗源[3] 杜启亮[3] 

机构地区：[1]广东工业大学自动化科学与工程学院,广东广州510090 [2]广东教育学院,广东广州510303 [3]华南理工大学自动化科学与工程学院,广东广州510640

出　　处：《中国石油大学学报（自然科学版）》2009年第5期159-163,共5页Journal of China University of Petroleum(Edition of Natural Science)

基　　金：广东省科技厅科技攻关项目(2005B10201006);广州市科技攻关引导项目(2003Z3-D0091)

摘　　要：利用粗糙集(RS)对不精确数据的处理能力,生成分类数据的边界集,替代原始样本作为训练集,减少训练集与获取的支持向量的数量,然后使用支持向量机的最小序列优化(SMO)算法改进回归学习机的性能。将粗糙集与SMO回归算法结合提出一种混合函数回归算法RS-SMO-RA。在常用SMO回归算法SMO-RA基础上,扩增一段简短的生成边界样本的算法程序。仿真结果表明,算法RS-SMO-RA的效率更高,且能够改进学习结果的性能。Rough set （RS） was utilized to analyze imprecise data and get the boundary set of the classified data. The boundary set can substitute the original inputs as a training subset, and the size of the training set and the gained support vectors are shorten. Then, the learning machine has solutions with high quality by sequential minimal optimization （SMO） algorithm of regression. Based on rough set and SMO algorithm of regression, a hybrid algorithm of RS-SMO-RA was presented for the enhanced capability of machine learning. For differentiating boundary samples, a simple and short module was added to the common algorithm of SMO regression, SMO-RA. The presented RS-SMO-RA algorithm is verified with high efficiency and performance.

关 键 词：支持向量回归机 SMO回归算法 边界样本集 粗糙集 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]