Research On Social Networking Personalized Recommendation Technology

Web2.0technology led the word into a Social Network Services (SNS) era. Social networks,such as China’s Sina Weibo, Facebook, MySpace, Twitter etchas become the most influentialplatforms.Combined user group and information, SNS not only allows users to quickly and easilyaccess and share information, but also expanding the user’s social scope.Since themost of theinformation in social networks spread along the user’s friends relationship, the friend relationshiphas become an very important part of social networking sites.Exploiting the socialnetworking,inbasis of offline relationships,users established corresponding online relationships,and slowly beganto produce a simple online friend relationships, this simple online friends largely make uptheemotional emptiness in modern relationship.The potential friends recommendation function is avery practical and popular social networking service which can help users establish a goodrelationship between social network friends group more quickly and make they more rapidlyintegrate intothe SNS.Currently,there are many social networking personalized recommendation methods, thesemethods are generally dependent on the topological structure of the social networks or user’spersonal data,but those static information do not have timeliness. In this paper,we proposed animproved collaborative filtering method to generate the list of potential friends for current user.Ouralgorithm combined genetic algorithm to determin user’s interest groups,and then improved theuser similarity degree calculation algorithm，provides a kind of user’s interest change adapting andbilateralinterest recommendation algorithm.For those new users who has just join in the social network,since they don’t have any socialingrecord yet, we designed interest-based group experts recommending approach to solve thisproblem. Firstly, our algorithm take advantage of Google PageRank algorithm to calculate eachuser’s "reputation"in the group, those who have a high reputation value are expert users.Thengenerate recommendations for new users according to the expert users’ experience. Finally,experimentalresults demonstratethesuperiority ofourmethod.