摘要
针对层次结构对二叉树支持向量机分类性能影响较大的问题,提出了一种改进的完全二叉树支持向量机构建方法。基于帕累托原则以核心圈样本最近类间距离和类内计算半径圈样本平均密度建立了类间差异性估计策略,将类间距离大且类内样本分布紧密的类别最先分离出来,并提出了构建完全二叉树的算法步骤。通过在UCI标准数据集上与其他SVM多类分类算法作比较,验证了改进算法的优越性。以TE过程故障诊断为研究对象,基于核主成分分析提取故障特征,应用改进的二叉树支持向量机实现了故障的准确识别。
This paper presents an improved complete binary tree SVM construction method,according to the strong influenc e of tree hierarchy on binary tree SVM classifier performance. The estimate measure of otherness between different classes was constructed by the nearest distance of samples of kernel-circle between different classes and the distance of samples of calculating-radius-circle in one class based on Pareto principle,so the class which has bigger distance from other classes and closer sample distribution within itself was first separated. The algorithm steps of constructing complete binary tree was proposed. Compared with other SVM algorithm on standard UCI data sets,the superiority of improved algorithm is verified. Fault diagnosis of TE process being taken as the research object,feature vectors of faults were extracted based on kernel principal component analysis,and the faults were diagnosed by the improved binary tree SVM accurately.
作者
陈柏志
石宇强
詹钧凯
邬江波
CHEN Baizhi;SHI Yuqiang;ZHAN Junkai;WU Jiangbo(School of Manufacturing Science and Engineering,Southwest University of Science and Technology,Mianyang 621010,Sichuan,China)
出处
《西南科技大学学报》
CAS
2018年第2期101-107,共7页
Journal of Southwest University of Science and Technology
关键词
支持向量机
二叉树
哈夫曼树
TE过程
故障诊断
Support vector machine
Binary tree
Huffman tree
Tennessee Eastman process
Fault diagnosis
