Research On Interest-based Collaborative Filtering Recommender System In P2P Network

With the wide application of P2P networks and increasing of information sharing, P2P network provides not only the information resources but also a great platform for exchanging information and sharing information. Getting valuable information from enormoous stockpile of information become more and more difficult. How to make full use of this information exchange platform is the challenge we faced. Information Recommendation makes that efficiently sharing information between people who have common interest become possible.Considering the characteristics of P2P networks, a collaborative filtering recommendation system based on interest-domain is proposed in this paper. In order to avoid flooding that exist in P2P networks, and improve the efficiency of the algorithm ,we use a method of incremental calculation of similarity based on interest-domain to calculate the similarity. Taking into account privacy issues, a mechanism that allows the user to choose whether rates can be shared with other users is added. In the scheme, according to the query logs of the user, it extracts the users' interests. Each node in the system maintains a user profile and friends list. Description file of user interest determines which interest-domain the node belongs to. The node adds the nodes with the similar interest to Friend List. Then uses an incremental method to calculate interest similarity, at last selects Top-N items recommended to user. Experimental results show that the scheme proposed in this paper can solve the data privacy, as well as the algorithm scalability issues.Considering the self-organizing of node and the existence of malicious nodes, we add the element of trust in the recommend system in P2P environment, using recommend error as the trust value between nodes. The nodes that most trust construct the friend list. Experimental results show that, the algorithm proposed in this paper has smaller MAE value than the traditional algorithm.