摘要
针对一般最小二乘支持向量机处理大规模数据集会出现训练速度慢、计算量大、不易在线训练的缺点,将修正后的遗忘因子矩形窗方法与支持向量机相结合,提出一种基于改进的遗忘因子矩形窗算法的在线最小二乘支持向量机回归算法,既突出了当前窗口数据的作用,又考虑了历史数据的影响.所提出的算法可减少计算量,提高在线辨识精度.仿真算例表明了该方法的有效性.
Aiming at the problem that the large-scale samples training process is slow and large computation, and difficult to train online for the standard least squares support vector machines, a learning algorithm of online least squares support vector machines regression（OLS-SVMR） based on improved rectangular window with forgetting factor（IRWFF） method is proposed by combining the modified rectangular window with forgetting factor algorithm with support vector machines. The present and past window data are considered simultaneously. The proposed algorithm has less computation and high accuracy. The simulation results show effectiveness of the algorithm.
出处
《控制与决策》
EI
CSCD
北大核心
2009年第1期145-148,共4页
Control and Decision
基金
国家自然科学基金重点项目(60534020)
国家973计划项目(2002CB312205)
高等学校博士学科点专项科研基金项目(20070006060)
北京市重点学科基金项目(XK100060526)
