The Research Of Friend Recommendation In Social Network Services

With the rapid development of Web2.0technology, the global Internet gradually goes into the era of SNS (Social Network Services). The representative social network sites have become a powerful information platform, such as Facebook, Twitter, MySpace and Flickr, etc. They combine with user groups and information, then users can conveniently get and share information, and greatly expand the social groups.Friend relationship of the user groups, as one of the transmission way of reliable information on the social network site, is a very important part of SNS. During the era of social network, many users transfer the offline friendships to the online relationships, and even generate singly online relationships unrelated to the offline friendships. Friend recommendation is one popular and bread-and-butter personalized service of social network to help users, especially the new users, to establish a good friend circle and to easily integrate into the entire social network information services.At present, many friend recommendation methods are either based on the topological structures of social network or derived from profile information of users, however, they are static and don’t have the character of timeliness. So as for the new users in the SNS, this paper focuses on the friend recommendation method of user interest circles partition which includes hard partition and soft partition. In view of the hard partition will enable the user to belong to just one interest circle, which will lead to recommend the users just only come from one interest circle and ignores the user’s other hobbies information, this paper adopts a method based on fuzzy spectral clustering algorithm to solve the interest circle hard partition problem and uses the GKA algorithm to ensure its global partition results to improve the efficiency of friend recommendation. It introduces in detail the design idea and design flow of fuzzy spectral clustering algorithm. In this paper, we use the social tags as the user character. It will complete data preprocessing and modeling on user tags as the experimental sample set. We reform the scaling exponent which should be manually set, and adopt a method of automatic setting, reducing the parameter’s randomness. And then it describes in detail the implementation process of friend recommendation based on spectral algorithm and fuzzy spectral algorithm. Through the precision and recall indexes, we verify this two algorithm’s results.Secondly, in view of the problem of new user cold start, namely, how to recommend friends for new users, this paper brings forward a method based on user data import which will access to the related information about this user in other products. This method is very suitable for enterprise. At last, taking the configuration of LDAP server, OAuth authentication and accessing to the user’s e-mail information for example, this paper introduces the whole realized process about this method and verifies the validity of the method.