摘要: 针对多工况对象的单模型建模中存在的回归精度差和泛化能力弱的问题,提出基于仿射传播聚类的LS-SVM多模型建模方法。该方法用仿射传播聚类算法对样本进行聚类,采用LS-SVM的方法对子类样本分别建立模型。测试样本根据相似性的测度进行归类,并用所属子类的模型进行预测输出。将该建模方法用在丙烯浓度的软测量建模实验中,结果表明该方法有较高的回归精度和较好的泛化能力。
关键词:
多模型,
仿射传播聚类,
最小二乘支持向量机,
建模
Abstract: The single model of the object with multiple working positions usually suffer from bad accuracy. To solve the problem, a Least Squares-Support Vector Machine(LS-SVM) multi-model modeling method based on affinity propagation clustering is presented. In this method, affinity propagation clustering is used to cluster training samples. The sub-models are trained by LS-SVM. The predicted values of the test samples are estimated by the sub-models after it is classified by similarity measurement. The proposed method is applied for soft-sensing modeling to predict the propylene concentration. Experimental results indicate that the proposed method has a superior regression accuracy and good generalization ability.
Key words:
multi-model,
affinity propagation clustering,
Least Squares-Support Vector Machine(LS-SVM),
modeling
中图分类号:
宋坤, 李丽娟, 赵英凯. 基于AP的LS-SVM多模型建模算法[J]. 计算机工程, 2011, 37(14): 169-171.
SONG Kun, LI Li-Juan, DIAO Yang-Kai. LS-SVM Multi-model Modeling Algorithm Based on AP[J]. Computer Engineering, 2011, 37(14): 169-171.