Research Of Credible User Recommendation Model In Network Community Based On Interest Relationship

Since Google’s browser has been used, recommendation systems has brought us big convenience when we want to search information on the Internet. With the development of Internet, Social Networking Services (SNS) attracts the research focus of personalized recommendation. Due to more and more people use Social Networking Services, the users are difficult to find the information quickly which will be suitable for their own interests. So a lot of SNS websites start to provide personalized information recommendation service. But due to the limitations of traditional recommendation system and the vast number of users in social networks, it’s a great challenge to rapidly find users’ requirements.Based on the background of the above problems, the paper studies user recommendation method in the social network. At the beginning, the paper studies and summarizes the key technologies involved in the paper, mainly includes the personalized recommendation technology, social network analysis and trust evaluation method, etc.; And then the paper chooses sina weibo as the research object, and chooses attributes that can represent the characteristics of users’interests in the social network in sina weibo,including users’ friends list, users’labels list and users’description, to establish a model to evaluate users interests similarity and users interests similarity matrix. The paper then can get a network graph based on users’ interests relationship; The paper divides the graph of users’ interests realitionship into different interests cliques by using CPM method, an overlapping community division method, and builds the users interests community by combines the similar cliques. At the part of Screening trusted users within the users interests community, the paper references to Google’s PageRank algorithm,which is used to sort pages in search engine,and establishes user credibility evaluation model. The user’s credibility is divided into two dimensions:credibility of user’s attributes and credibility of others who focus on you. User’s global credibility reflects on the user’s own credibility and the credibility of user’s fans. Finally, by comparing others’ credibility with the target user’s in the same interests community, the paper chooses users with higher interests similarity and higher credibility as the target user’s recommendation list, recommends to the target user with high value of users as potential friends. The paper uses a piece of code that programmed by JAVA,a kind of programming language,to get the users information from the API of users data sets in sina weibo, and verifies the recommendation model of trusted users in the network community based on users interests relationship by comparing it with the traditional recommendation model based on social relationship. This paper chooses the user’s average interest similarity of the recommendation lists to measure the effectiveness of the models. Experiment results of this paper shows that the recommendation model proposed by the paper gets the recommendation lists with generally higher interest similarity than that of recommendation model based on social relations, the recommendation model puts forward by this paper could get the recommendation lists which meets the users’ interests better and improve the quality of the recommendation system.To verify the practical value of the users recommendation model established by this paper, the5th section of this paper randomly selects113users and sends each of the them two kinds of recommendation lists,each including5friends,and one of the lists is produced by the model established by this paper,and the other is produced by the traditional users recommendation model based on social relationships, and then the survey asks users some questions about their feelings about the recommendation lists to analyze the effects of two kinds of recommendation models. According to the survey, the recommendation model established by this paper can recommend more new users, and the user recognition is higher. And compared with social relationships, users will be more interested in the reasons of interest similarity and user credibility. What’s more, it’s much more possible for users to add the recommendation users as friends whose interest similarity and credibility are higher.When comes to the recommendation model based on social relationships, users will be more probably hope that someone can introduce the recommendation users to them.