| 特征抽取的一个新算法 |
| |
| 引用本文： | 梁逸曾,俞汝勤,董有方.特征抽取的一个新算法[J].湖南大学学报(自然科学版),1989,16(2). |
| |
| 作者姓名： | 梁逸曾 俞汝勤 董有方 |
| |
| 作者单位： | 湖南大学化学化工系
(梁逸曾，俞汝勤)，湖南省岳阳市电子研究所(董有方) |
| |
| 基金项目： | 国家自然科学基金资助项目 |
| |
| 摘 要： | 特征抽取无论在模式识别还是在光谱信道和色谱柱组合选择中,皆具有重要意义.本文提出了一个基于复相关系数的特征抽取新方法.藉扫描算法,本法可在微机上方便快迷运行.将该法对用于诊断两类疑难疾患的模式识别数据进行了特征变量的选择.获得满意结果.
|
| 关 键 词： | 化学计量学 特征抽取 复相关系数 |
A New Algorithm for Feature Selection |
| |
| Liang Yizeng Ru Yuqin.A New Algorithm for Feature Selection[J].Journal of Hunan University(Naturnal Science),1989,16(2). |
| |
| Authors: | Liang Yizeng Ru Yuqin |
| |
| Institution: | Liang Yizeng Ru Yuqin Department of Chemistry and Chemical Engineering
Dong Youfang Yueyang Institute of Electronics,Hunan |
| |
| Abstract: | The feature selection is very important for the pattern recognition as wll as for the optimal combination of spectal channels or chromatographic columns. Based on the multiple correlation coefficients, a new method for feature selection is developed in this paper,and with the help of the scanning algorithm, it can be realized easily and quickly on a micro-computer. Using this method, we select the feature variables in pattern recognition for diagnosis of two kinds of diseases and the satisfactory results are obtained. |
| |
| Keywords: | chemometrics feature selection multiple correlation coefficient |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《湖南大学学报(自然科学版)》浏览原始摘要信息 |
| 点击此处可从《湖南大学学报(自然科学版)》下载免费的PDF全文 |
