计算机科学与探索 ›› 2010, Vol. 4 ›› Issue (10): 899-908.DOI: 10.3778/j.issn.1673-9418.2010.10.004

• 学术研究 • 上一篇    下一篇

Co-Training——内容和链接的Web Spam检测方法*

魏小娟1,2+, 李翠平1,2, 陈 红1,2   

  1. 1. 中国人民大学 数据工程与知识工程国家教育部重点实验室, 北京 100872
    2. 中国人民大学 信息学院, 北京 100872
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2010-10-01 发布日期:2010-10-01
  • 通讯作者: 魏小娟

Content and Link Based Web Spam Detection with Co-Training*

WEI Xiaojuan1,2+, LI Cuiping1,2, CHEN Hong1,2   

  1. 1. Key Lab of Data Engineering and Knowledge Engineering of MOE, Renmin University of China, Beijing 100872, China
    2. School of Information, Renmin University of China, Beijing 100872, China
  • Received:1900-01-01 Revised:1900-01-01 Online:2010-10-01 Published:2010-10-01
  • Contact: WEI Xiaojuan

PDF

摘要： Web spam是指通过内容作弊和网页间链接作弊来欺骗搜索引擎, 从而提升自身搜索排名的作弊网页, 它干扰了搜索结果的准确性和相关性。提出基于Co-Training模型的Web spam检测方法, 使用了网页的两组相互独立的特征——基于内容的统计特征和基于网络图的链接特征, 分别建立两个独立的基本分类器; 使用Co-Training半监督式学习算法, 借助大量未标记数据来改善分类器质量。在WEBSPAM-UK2007数据集上的实验证明：算法改善了SVM分类器的效果。

关键词: Web spam检测方法, 内容作弊, 链接作弊, Co-Training算法

Abstract: Web spam attempts to deceive search engine by crafting the content of Web pages or creating tight knit community of links around irrelevant Web pages, for the purpose of getting an undeserved high rank. It maliciously influences the accuracy and relevancy of ranking algorithms. This paper proposes a novel Web spam detection method based on Co-Training model. It builds two basic classifiers separately considering link-based and content- based features, then leverages unlabeled data along with a few labeled examples to boost the performance of the classifier through a semi-supervised algorithm—— Co-Training model. And the experimental results on WEBSPAM- UK2007 dataset demonstrate that the algorithm improves the efficiency and accuracy of SVM classifier.

Key words: Web spam detection method, content-based spam, link-based spam, Co-Training

中图分类号: 

京ICP备13024263号-1
Copyright © 《计算机科学与探索》编辑部
技术支持：北京玛格泰克科技发展有限公司
总访问量
今日访问
在线人数