摘要
在标准支持向量回归在线学习的基础上,提出了一种加权支持向量回归在线学习方法(WOSVR),即加权支持向量机中针对不同样本点使用不同惩罚系数C,且不同惩罚系数C反映了样本重要性的不同,WOSVR中近期数据重要性大于历史数据重要性.使用基准数据Mackey-Glass混沌序列进行了相关验证实验.结果表明,加权支持向量回归在线学习方法能有效修改模型.
This paper investigated a weighted on-line on the on-line support vector regression. In weighted ter C is used variably with different samples, which support vector regression （WOSVR） approach based support vector regression, the regularization paramedenotes the different importance of the samples. In WOSVR, the importance of the recent past data is higher than that of the distant past data. Comparative tests were performed using chaotic Mackey-Glass benchmark. The experimental results show that the pro- posed method can change the model more effectively.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2009年第6期927-930,共4页
Journal of Shanghai Jiaotong University
基金
上海市科委科技创新行动计划项目(08dz1202502)
上海高校选拔培养优秀青年教师科研专项基金
上海海事大学科研基金
关键词
支持向量机
加权支持向量回归
在线学习
support vector machine
weighted support vector regression
on-line learning
