Research On The Key Issues In Personalized Recommendation Based On SNS

Since entering21st century, the websites have been socializedobviously. More and more websites integrate with shopping，makingfriends，chatting，forum and bolg, which could provide users with all-roundservice. As a way of personalized service, personalized recommendationsystem can recommend items which users interested and can help usersmake decisions. This system significantly reduces the cost of people gettingthe information.At present, the actual recommendation systems mostly adoptcollaborative filtering recommendation algorithms. The collaborativefiltering recommendation System exists the problems of data sparseness,cold starting, low expandability and so on. In the environment of socialnetworking, some researchers brought the social trust mechanism into thetraditional collaborative filtering recommendation System, so that theserecommendation Systems based on trust improved the above problemspartly.However, the trust between users can not be accurately described. Inmany cases, the useful effect of user trust in the recommendation is evenworse than the traditional methods used by the user similarity. Someresearchers have proposed the method of combining traditional trust-basedcollaborative filtering and trust-based recommendation, but these studiesneglect the effect of users groups characteristics on the recommendation.Focusing on how to improve the recommendation satisfaction of usersabout recommendation system, this paper explores and researches some keyissues of the personalized recommendation system in the environment ofsocial networking.Firstly, aiming at the shortage of present recommendation algorithms,this paper presents a new social recommendation model-Cliqueswalk using the analytical approach of social networking. Experiment resultsdemonstrate that the new method can greatly narrowed the scope ofsearching target rating information and the effect of recommendation isbetter than the existing methods of Collaborative filtering, trust-based andtheir combined.Secondly, taking into account of the different effect of different usersin social networking, this paper adopts the analytical approach of socialnetworking and Web Ming method, designs the search algorithm ofauthoritative users(opinion leader) and provides the correspondingrecommendation algorithm. Separating the authoritative users from theordinary users is not only improves the effectiveness of the recommendation,but also brought a better user experience.Finally, as a service in social networking service, blog system holds animportant position. This paper provides the user multiple interests of thetrust-based personalized recommendation algorithm. The algorithm dividesthe whole trust score into multi-trust scoring so that the trust relationshipcan better roundly measure the users’ interest, and then bring the combinedapproach of Collaborative filtering recommendation and trust-basedrecommendation into multi-interest recommendation. The experimentalresults show that the new algorithm not only has better recommendationeffect for the ordinary users, but also improves the recommendation effectfor the cold start users.