摘要
提出了一种基于贝叶斯证据框架的回归型支持向量机的参数选择和软测量建模方法。证据框架的第1层推理用来解释支持向量机的训练。证据框架的第2和第3层推理用于自动调整正则化参数和核参数并使其接近最优值。然后,将这种基于贝叶斯证据框架的回归型支持向量机用于估计聚丙烯腈生产过程的质量指标。仿真结果表明了该方法的有效性。
A parameter selection and soft sensor modeling method based on Regression support vector is presented within Bayesian evidence framework. The Level 1 inference of the evidence framework can interpret training of the SVM. And Levels 2 and 3 of the evidence framework allow automatic adjustment of the regulation parameter and the kernel parameter to their near-optimal values. Moreover, the regression support vector based on Bayesian evidence framework is used to estimate the quality figure of polyacrylonitrile productive process. Simulation results indicate the effectiveness of the proposed method.
出处
《南京航空航天大学学报》
EI
CAS
CSCD
北大核心
2006年第B07期136-138,共3页
Journal of Nanjing University of Aeronautics & Astronautics
关键词
证据框架
贝叶斯
软测量
支持向量机
evidence framework
Bayesian
soft-sensor
support vector machine