摘要
支持向量回归问题的研究,对函数拟合(回归逼近)具有重要的理论和应用意义。借鉴分类问题的有效算法,将其推广到回归问题中来,针对用于分类问题的SOR支持向量机有效算法,提出了SORR支持向量回归算法。在若干不同维数的数据集上,对SORR算法、ASVR算法和LibSVM算法进行数值试验,并进行比较分析。数值实验结果表明,SORR算法是有效的,与当前流行的支持向量机回归算法相比,在回归精度和学习速度上都有一定的优势。
The research on support vector regression has an important theoretical and applicable significance on function regression（re-gression approximation）.Using effective arithmetic of classifier for reference,it is extended to the matter of regression.Aimed at the effective SOR（successive overrelaxation for support vector） arithmetic for classification,the effective SORR（successive overrelaxation for support vector regression） arithmetic is proposed.On several different data aggregation of dimensions,the numerical experiments and comparison are carried out on SORR arithmetic,ASVR arithmetic and LibSVM arithmetic.The numerical experiments show that the SORR arithmetic is effective and it has certain advantages on learning speed and regression accuracy,compared with the current popu-lar support vector regression arithmetic.
出处
《计算机工程与设计》
CSCD
北大核心
2008年第6期1489-1490,1493,共3页
Computer Engineering and Design
基金
国家自然科学基金项目(10571109)
山东省自然科学基金项目(2007ZRB019FK)
关键词
连续超松弛支持向量
连续超松弛支持向量回归算法
函数拟合
学习速度
回归精度
successive overrelaxation for support vector
successive overrelaxation for support vector regression arithmetic
function regression
learning speed
regression accuracy
