| 利用未标记数据提高SVM分类器性能的研究 |
| |
| 引用本文： | 祝宇,聂峰光,郭力. 利用未标记数据提高SVM分类器性能的研究[J]. 计算机工程与应用, 2006, 42(27): 166-167,170 |
| |
| 作者姓名： | 祝宇 聂峰光 郭力 |
| |
| 作者单位： | 中国科学院过程工程研究所多相反应实验室,北京,100080;中国科学院研究生院,北京,100049;中国科学院过程工程研究所多相反应实验室,北京,100080 |
| |
| 摘 要： | 监督学习算法的一个主要困难在于需要大量标记过的训练集数据,采用人工的方法不够现实。文章提出了SVM分类器在少量标记训练样本情况下,采用Rocchio法和KNN方法从大量的未标记数据中,挑选相似度较高、区别度较大的数据加入到训练集中,弥补训练样本的不足。实验表明该算法有效地利用了丰富的未标记数据,减少了人工标记量,较好地提高了SVM分类器的性能。
|
| 关 键 词： | 文本分类 未标记 Rocchio法 K近邻法 支持向量机 |
| 文章编号： | 1002-8331-（2006）27-0166-02 |
| 收稿时间： | 2005-12-01 |
| 修稿时间： | 2005-12-01 |
Research on Using Unlabeled Data to Improve the Performance of SVM Classifier |
| |
| ZHU Yu,NIE Feng-guang,GUO Li. Research on Using Unlabeled Data to Improve the Performance of SVM Classifier[J]. Computer Engineering and Applications, 2006, 42(27): 166-167,170 |
| |
| Authors: | ZHU Yu NIE Feng-guang GUO Li |
| |
| Affiliation: | 1.Lab of Multi-Phase Reaction,Institute of Process Engineering, Chinese Academy of Sciences,Beijing 100080; 2.Graduate University of Chinese Academy of Sciences,Beijing 100049 |
| |
| Abstract: | The trouble of supervision method is the demand of much labeled data for training.It's difficult by human inspection.This paper presents a method to use the unlabeled data to improve the performance of SVM classifier,which has only a few labeled training examples.It selects some data with high similarity and appropriate difference from the unlabeled data set by the algorithm of Rocchio or K Nearest Neighbor.Experiments show that this method can use the unlabeled data effectively,reduce the need of labeled data,and improve the classifier's performance. |
| |
| Keywords: | text categorization unlabeled Rocchio's algorithm K Nearest Neighbor Support Vector Machine |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
