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.
出处
《中国石油大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
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)
