Due to the rapid development and wide applications, Internet has led to the information explosion on the WEB. It becomes research hotspots at present how to design effective algorithms and systems based on the technology of WEB Usage Mining. In this paper, we develop a recommendation system called WRS (Web Recommendation System), which is based on the application of WEB. In WRS, we propose a novel algorithm that makes use of the technology of image segmentation to cluster access modules, and adopts the parallel longest common subsequence algorithm to discern users' behaviors. Theoretical analysis and laboratory result show that our system is more effective and the recommendation performance is improved after using the new method.