用半监督聚类算法实现WEB文本挖掘

随着Internet网络的高速发展,海量的未标签文档和相对少量的已标签文档是当前Web文档的一个普遍情形,如何有效的利用少量的已标签文档去聚类海量的未标签文档,从而更好地获取有价值的信息,即半监督学习问题,已成为当前研究的热点.本文针对目前Web文本挖掘领域的无监督学习算法的检测率不高,而监督学习算法需要大量的标签数据又不易获得的问题,将半监督中的标签绑定技术与优化球形k-均值聚类算法相结合进行Web文本挖掘,并使用真实的测试数据对Web文本挖掘系统进行实验.结果表明本文方法对有价值文本具有较高检测率及较低的误报率,整体检测性能优于基于监督和无监督学习的Web文本挖掘算法.

Semi-supervised Clustering Method Based on Web Text Mining

With the rapid development of Internet network,magnanimous data that being not tag or less tag data is a universal situation in present Web documents.How to get useful information better using a small amount of data that have been tagged to clustering mass of not tag data,that is Semi-supervised learning question,has been a hot point in research today.in view of low examination rate in web text mining＇s non-supervised learning and a lot of tag data being not easily available in supervised learning algorithm,therefor tag binding of semi-supervised learning algorithm and refined bisecting k-means clustering algorithm will be combined for Web text mining.Finally real examination data for experiment shows that this mathod has higher examination rate and lower misinformation rate for texts on the value.It＇s overall detection performance is superior to the web text mining algorithms based on non-supervised learning and the supervision learning.