| 基于样本类可分性分析的特征选择研究 |
| |
| 引用本文： | 崔建新,洪文学,高海波,王金甲.基于样本类可分性分析的特征选择研究[J].燕山大学学报,2008,32(6). |
| |
| 作者姓名： | 崔建新 洪文学 高海波 王金甲 |
| |
| 作者单位： | 1. 燕山大学,电气工程学院,河北,秦皇岛,066004;河北省测试计量技术及仪器重点实验室,河北,秦皇岛,066004
2. 燕山大学,电气工程学院,河北,秦皇岛,066004 |
| |
| 基金项目： | 国家自然科学基金资助项目
|
| |
| 摘 要： | 在传统类间散布矩阵理论的基础上,提出了类间的两两散布矩阵和类间重叠系数矩阵.传统的类间散布矩阵对于两类或多类的类别均值和全局均值之间距离值相近时难以区分,而且对于方差大而分类信息差的向量也无能为力.类间重叠系数矩阵可以剔除方差大而分类信息差的向量,两两类间散布矩阵则用于区分类别均值和全局均值之间距离值相近的向量.实验证明该方法生成的特征向量取得的分类效果较好.
|
| 关 键 词： | 多元信息 散布矩阵 类间重叠系数 特征选择 |
Study on feature selection based on sample sort separablity analysis |
| |
| CUI Jian-xin,HONG Wen-xue,GAO Hai-bo,WANG Jin-jia.Study on feature selection based on sample sort separablity analysis[J].Journal of Yanshan University,2008,32(6). |
| |
| Authors: | CUI Jian-xin HONG Wen-xue GAO Hai-bo WANG Jin-jia |
| |
| Institution: | CUI Jian-xin1,2,HONG Wen-xue1,GAO Hai-bo1,WANG Jin-jia1 (1. College of Electrical Engineering,Yanshan University,Qinhuangdao,Hebei 066004,China,2. Measurement Technology , Instrumentation Key Lab of Hebei Province,China) |
| |
| Abstract: | On the basis of traditional sorted scatter matrix theory, the two-two sorted scatter matrix and sorted overlap coefficient matrix were presented. It's difficult for traditional sorted scatter matrix to separate the two-sorted or multi-sorted samples when the distance of the sorted mean and the whole mean is close. It is also helpless to the variable that has bigger variance and has little effect on classification. Sorted overlap coefficient matrix can eliminate the variable that has bigger variance and has ... |
| |
| Keywords: | multivariate information scatter matrix sorted overlap coefficient feature selection |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
