| 改进的SVM算法及其在故障诊断中的应用研究 |
| |
| 引用本文： | 张金泽,单甘霖.改进的SVM算法及其在故障诊断中的应用研究[J].电光与控制,2006,13(6):97-100. |
| |
| 作者姓名： | 张金泽 单甘霖 |
| |
| 作者单位： | 军械工程学院光学与电子工程系,石家庄,050003 |
| |
| 摘 要： | 介绍了支持向量机用于解决模式分类问题的基本原理，在对传统的多分类方法OVO（one—versus—one）深入分析的基础上，针对其存在的不可分类区问题，提出了一种改进的模式分类方法KSVM（KNN—SVM），将k-近邻方法嵌入到SVM算法中解决不可分类区问题，进一步提高了分类准确率。应用KSVM分类方击进行模拟电路的故障诊断，实验结果验证了该方法的有效性和实用性。
|
| 关 键 词： | 支持向量机(SVM) 多类分类方法 k-近邻法 故障诊断 |
| 文章编号： | 1671-637X（2006）06-0097-04 |
| 收稿时间： | 2005-07-01 |
| 修稿时间： | 2005-07-012005-07-13 |
Improved SVM algorithm and its application in fault diagnosis |
| |
| ZHANG Jin-ze,SHAN Gan-lin.Improved SVM algorithm and its application in fault diagnosis[J].Electronics Optics & Control,2006,13(6):97-100. |
| |
| Authors: | ZHANG Jin-ze SHAN Gan-lin |
| |
| Affiliation: | Department of Optics and Electronics Engineering, Ordnance Engineering College, Shijiazhuang 050003, china |
| |
| Abstract: | The basic principle of Support Vector Machines used in Pattern Classification is simply introduced.The traditional one-versus-one(OVO) method for multi-class classification is analyzed.To overcome the unclassifiable region problem existed inherently in one-versus-one method,an improved method of KSVM(KNN-SVM) is presented for multi-class SVMs,in which K-Nearest Neighbor(KNN) is embedded in SVM algorithm for solving the unclassifiable region problem,the classification accuracy is improved further.The application of KSVM in analog circuit fault diagnosis proved the validity and practicability of the new method. |
| |
| Keywords: | support vector machines(SVM) multi-class classification k-Nearest Neighbor fault diagnosis |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
