Peer-to-peer usage analysis in dynamic databases

With the rapid growth of Internet and computer techniques, the huge number of information is thus available to be analyzed for modeling user behaviors. Peer-to-peer architecture provides the large-scale community behaviors for information exchanging and sharing. Usage behaviors can be defined as the sequential order as the requests or downloads performed on each node in P2P system. Sequential pattern mining (SPM) can be used to discover usage behaviors to facilitate efficient decision-making. In the past, the fast updated sequential pattern (FUSP)-tree structure was proposed for handling sequence insertion and sequence deletion without candidate generation. Transaction modification is, however, also an important issue in real-world applications. In this paper, a maintenance (FUSP-TREE-MOD) algorithm to efficient update FUSP-trees for sequence modification in dynamic databases is proposed. The proposed approach can thus enhance behaviors modeling in dynamic P2P system for extracting sequential patterns or relationships occurring in a large number of nodes. Experimental results indicate that the proposed algorithm outperforms batch approaches in maintaining discovered sequential patterns.