摘要
为提高支持向量回归在时间序列预测应用中的学习速度和泛化性能,提出了稀疏型支持向量回归方法.通过牛顿优化法,直接优化支持向量回归的原始问题.然后利用Cholesky分解更新原始优化中的Hessian矩阵实现稀疏型支持向量回归算法.最后将该算法运用到Mackey-Glass,Lorenz和Logistic混沌时间序列预测,仿真结果表明本文提出的方法能够在确保预测精度的前提下,有效地降低支持向量的个数.
Sparse support vector regression (SpSVR) method is proposed to improve the leaning speed and generalization performance in time series prediction. Firstly, the primal problem of support vector regression (SVR) is directly optimized through Newton optimization method. Then, in order to realize the sparseness of SVR, Cholesky decomposition is used to update the Hessian matrix in SVR primal problem. Finally, the proposed algorithm is applied to Mackey-Glass, Lorenz and Logistic chaotic time series predictions. The simulation results indicate that the SpSVR is able to effectively reduce the number of support vectors with guaranteed prediction precision.
出处
《系统工程学报》
CSCD
北大核心
2011年第5期584-591,共8页
Journal of Systems Engineering
基金
国家空中交通管制委员会基金资助项目(GKG200902004)
关键词
时间序列
支持向量回归
稀疏
牛顿优化法
time series
support vector machines
sparse
Newton optimization