Research On P2P Search Mechanism Based On User Interest

The development of Peer-to-Peer network technology and the growth of network resources not only bring the users in great difficulties in searching, locating and retrieving resource, but also can't meet the users' personalized requirements increasingly.The search scope of P2P-based search engine would include all the information resources opened on the network in theory. Therefore, the level of P2P search engine's development determines the utilization rate of Internet resources in some way. From the user interest point of view, this paper researches the model of personalized P2P search engine system and the specific search algorithm.Based on the analysis of existing P2P search technology features, this paper introduces Super Peer as system "server",introduces Agent technology as the basis for P2P search engine system modeling, constructs a P2P search engine system model and designs the system workflow. The introduction of Agent not only makes node discovery more flexible and comprehensive, but also improves the search efficiency and achieves the purpose of search within all kinds of different resources. This system model can picks out the information meeting user interest in the search results, and actively acquires user interest based on historical search.Searching for the resources which are associated with the user interests and satisfaction simply and efficiently is another focus of this paper. This paper mines user interests in two points of view:users and the contents. Then a P2P search method based on query expansion and node aggregation is proposed. Using the users'historical search behaviors, this method discovers the relationships of keywords and nodes, records them in each peer and updates continuously. In the future search, keywords relationship can expand search query to increase hit goals; nodes aggregation can divide interest category to narrow the search scope. This method can find the user interest and learn to update automatically, improve search performance and save users' time. The final simulation experiment proves that this method has higher search hit rate, lower nodes' load and network overhead.