Novel approach to system structure identification

作者：Su Yingying^*; Li Taifu; Cui Tianxiang; Nie Ling

来源：Journal of Information and Computational Science, 2008, 5(4): 1587-1595.

• Generalization abilities
• Inner structures
• Kernel methods
• Kernel Principal Component analyses
• Kernel principal component analysis
• Nonlinear components
• Nonlinear system identifications
• Novel structures
• Order
• Sample spaces
• Simulation results
• Structure identification
• Structure identifications
• Structure risk minimizations
• Support vector regression
• Support vector regressions
• System structures

Structure identification is one of the most difficult problems in nonlinear system identification including the consideration of order number and choice of model structure, which deeply impacts the accuracy and generalization ability of model. Aiming at the problem, a novel structure identification approach based on Kernel Methods-KPCA (Kernel Principal Component Analysis) and SVR (Support Vector Regression) is presented here. Firstly, the nonlinear components of sample space are extracted by KPCA, which confirms the order of sample space. Further, SVR with SRM (Structure Risk Minimization) is utilized to optimize the inner structure. On the basis of three examples, simulation results reveal that KPCA-SVR is an effective approach in solving nonlinear system structure identification. ? 2008 by Binary Information Press.

• 出版日期2008