基于多视图典型相关分析的垃圾网页检测

发布时间：2018-10-12 14:49

【摘要】：首先将垃圾网页特征分为两个不同的视图,即基于内容特征的视图和基于链接特征的视图,利用典型相关分析及其相关改进方法进行特征提取,生成两组新的特征;再对新生成的两视图特征采用不同组合方式产生单视图数据,并用这组数据作为训练数据构建分类算法。实验结果表明,将垃圾网页看成两视图数据,并应用多视图典型相关分析技术,可有效提高垃圾网页的识别精度。
[Abstract]:Firstly, the features of spam pages are divided into two different views, one is based on content feature and the other is based on link feature, and then the feature extraction is done by using canonical correlation analysis and its improved method to generate two new sets of features. Then, the new two view features are combined to generate single view data, and the classification algorithm is constructed by using this set of data as training data. The experimental results show that the identification accuracy of garbage pages can be effectively improved by using multi-view canonical correlation analysis technology and considering garbage pages as two-view data.
【作者单位】：山东师范大学信息科学与工程学院;山东省分布式计算机软件新技术重点实验室;
【基金】：国家自然科学基金资助项目(61170145) 国家教育部高等学校博士点专项基金资助项目(20113704110001) 山东省自然科学基金和科技攻关计划资助项目(ZR2010FM021,2008B0026,2010G0020115)
【分类号】：TP393.092

【参考文献】