| The openness, interaction and sharing of social network(SN) help it won the majority of Internet users, which becomes the main platform for Internet users to expressing ideas, information exchanging, and building social circles. FOAF (Friend of a Friend) is one the most famous SN project. By giving a full review of recent work about FOAF user analysis, we do the research from the following aspects: users’information extraction, users’interest modeling, products recommendation and profiles’authorization/privacy. To solve the following problems such as:how to mine the true features and interests of social network users; how to reason their information based on users’profile; how to recommend based on reasoning; how to judge the credibility of users, this paper proposed some basic technical methods and theories. The contents are shown as follows:1. Extremely dispersed customer relationship information maps and space storage dispersion make it is a challenge to effective mine and analysis the FOAF information. By integrating multiple online video network resources and multiple social network platforms, we proposed VRP to recommend latest online TV and movie programs for users. The data set comes from various online videos Web sites, and users’FOAF files are all stored in different databases and different computers. We applied FOAF and DBlink (database links) to recognize user entities and mine users’interest. In order to ensure maximum privacy of users’ information, we put user profiles in their local machines. It is a limitation only considering one user’s individual viewing historical information on traditional video recommendation systems, so we considered the same information of users’ friends, which could help us to achieve effective personalized recommendations.2. Our second phrase research, we mashed up Web information fusion and FOAF to design a belief reasoning recommendation system for user service (BRRUS). It took into account the spatial and temporal information in the process of user services, and it analyzed the relationships between users and information publishers based on their FOAF profiles. BRRUS used Markov chain Monte Carlo algorithm to improve the trust level of no registering users. This work also proposed an algorithm to detect particular experts, according to the spatial and temporal queries users submitted. These theoretical findings were supported by experiments on several test collections.3. Increasing development of Web2.0makes it possible to create personalized services to all Internet users, but for now in the existing network, there is only a single one-to-one mapping mode between Web users and Web data sources. The potential information among Web data sources and users services is not fully utilized, and it is important for next generation network to improve customer satisfaction by ensuring the credibility of the whole network data. So, in the third phrase research, by integrating Web sources and predicting Web users’potential relationship of resources based on evolutionary game theory and FOAF, we proposed a method named Network Source Trust Distinguish (NSTD) to make it possible. Experimental results show encouraging performance of the proposed NSTD model by studying cases of trusted network sources apply in different areas.4. At last, we make user personalized recommendation in a real world digital library, K-Gray Engineering Pathway (EP). Based on the users’education background, EP can give them different search results when the input the same the query by users’active behavior, who know their own education background and set the search area themselves, which make EP search result with high correction, but also leads to potential information loosing, for users have no idea of these knowledge exiting, for their limited educational. User also can get all search results from the homepage, independent of education background, but it made low correction and lots of time wasting to find what they really needed. So we focused on giving different and personal search recommendation for users with different educational level and accomplish this function automatically. For data, we research semantic relationships among knowledge, and then we classify them and establish the knowledge relationships model. For users, we can set up user profile by user log, then classify them and establish user model. When a frequent user come to EP looking for something, we can give information directly related and recommend knowledge not directly related but can arouse their interest, based on these two models. The experiments shows we make EP a more excellent expert who know users well enough to guide them, according to the statistic information such as education background, and by improving the collaborative recommend results in EP, our users can make the best use of their time by EP learning.Based on the integration of social network information, this study provides different recommendation theoretical systems. They can used to fuse the users information from scalable social network, to achieve a personalized recommendation for users with spatial and temporal information, and to determine the trust relationship amongst users to serve personalized recommendations. At last, we applied one of these methods to a real world digital library. |
PDF Full Text Request