Research On Recommendation System Based On Social Network Trust Relationship

The increasing popularity of social networks has prompted more and more users to like to get the content,shopping and information exchanges that they are interested in online.In order to adapt to the development of the times and give users a better experience effect,the recommendation system came into being,and has been continuously developed and improved in theoretical research and practical applications.With the increase of social network platforms and online users,the trust issues brought by different users in the recommendation process have gradually emerged.The derived trust perception recommendation system has gradually attracted widespread attention and became a research hotspot.The recommendation system based on trust relationships can provide important reference opinions when recommending items or projects for target users,and it is also easier to be accepted and loved.At present,the recommendations based on trust relationships are mainly based on the premise of more users with more prior information and comments,but often ignore the following three situations.1)New users lack prior information;2)There are few direct interactive information among most users;3)Most users may comply with the views and ideas of the leaders in the actual situation.In response to the above issues,in-depth research on this thesis,and the relevant solution was proposed.The specific work is as follows:(1)In order to be able to consider the trust relationship between users and specific behaviors more accurately and comprehensively,this thesis proposes a dynamic recommendation algorithm that integrates trust relationships and user behavior.The algorithm is based on the global trust and local trust in the trust relationship.After the time attenuation,the dynamic trust model is generated.At the same time,the consistency of user scoring information in social networks and the reliability between users to improve the credibility of the recommendation effect.The final generated prediction score was experimented on the related real public dataset.The experimental results showed that the dynamic recommendation algorithm that integrates trust relationships and user behavior characteristics is better than other similar algorithms in terms of recall and accuracy.Analyze this chapter algorithm in the film recommendation system,and provide reference for the improvement of the film recommendation.(2)Practice has shown that with the development of social networks,more and more users choose to follow the decision of opinion leaders.Therefore,this thesis proposes a method of opinion leader identification method to select the selection of opinion leaders for the user’s professionalism,trust,activity,and welcome;The final preference as a non-demand recommendation as a leader.Based on the above studies,this thesis proposes a social recommendation algorithm based on opinion leadership mechanism and user preference.The recommendation list generated by the generation after the scoring prediction is recommended to users.Experimental analysis on the related real public datasets shows that this algorithm has decreased in terms of the average and average absolute error compared with other comparison algorithms.Taking the Douyin short video as the research object,applying this chapter algorithm to short video recommendation can effectively improve the recommendation effect.There are 43 figures,5 tables,and 82 references in this thesis.