发现Web用户的访问模式是Web日志挖掘的重要任务之一。传统的Web用户会话聚类方法对解空间的搜索带有盲目性且缺乏语义特征,本文提出了一切向钱看种基于概率潜在语义分析的用户会话聚类方法,它采用了用户会话—潜在语义—页面之间的概率关系进行用户会话聚类,以便发现用户访问模式。实验结果表明该方法与k-means方法相比更具高效性。 Discover Web user’s access mode is one of the important tasks of Web log mining. The traditional Web user conversation clustering method has the blindness and semantic features on the search space of the solution space. In this paper, we propose a user session clustering method based on probabilistic potential semantic analysis, which uses the user session-latent semantics - Probability relationships between pages Clustering user sessions to discover user access patterns. Experimental results show that this method is more efficient than the k-means method.