基于超图聚类的用户行为模式挖掘

Mining User's Behavioral Patterns Based on Hypergraph Clustering

为解决代理服务器端用户行为日志的高维特征,提出一种基于超图模型的Web会话聚类方法。该方法利用页面归约和隐含在会话中的时间等统计信息描述会话,并在此基础上利用基于关联规则的超图对用户会话进行聚类,可以在较高层次上发现代理服务器端的用户行为模式。实验结果表明,该方法可以在大型的代理日志文件集中挖掘出有意义的用户行为模式,有效揭示用户的行为规律。

This paper presents a method to find user＇s behavioral patterns based on clustering web sessions using hypergraph. The method describes sessions using statistic information by reducing page URL and extracting implied time information from proxy log,and clusters user＇s sessions using hypergraph model based on association rules. Experimental results show that this method can find user＇s behavioral patterns from large proxy log,and effectively explore the law of user＇s regular behavior.