Research On The Friend Recommendation Method Based On Social Network

Since entering the Web2.0 era,social networking websites such as Facebook,Weibo and Twitter have been rapidly popularized.Users establish and expand their social circle that is in accordance with interests and preferences by the form of adding and following to realize the communication and information sharing among friends.As the number of users on social networking websites increased dramatically,it becomes increasingly difficult for users to find like-minded friends.In order to promote the user experience in social networking websites,friend recommendation system has gradually become an important part of the social networking websites.Friend recommendation method based on social network is becoming a hot research topic in the field of recommendation systems.Friend recommendation system based on social network can take the initiative to recommend new potential friends for users.In addition,it can effectively expand the scale of the user's social circle and improve user social experience,which has received the widespread attention.The traditional friend recommendation is mainly mining the user's interest preference and calculating the similarity among users to make good friend recommendations.After the social recommendation becomes a hotspot,friend recommendation also turns to the research on the relationship between users,and puts forward methods such as link prediction.Considering that user's interest and relationship are important factors for users,this paper mainly researches the hybrid friends recommendation method based on social network.The main innovation works are as follows:(1)A friend recommendation method based on user's potential features in social network is proposed.Firstly,this method use the latent semantic model to mining the implicit attribute features of the user through the user's behavior record and build the user's preference model about interesting and dating.Secondly,we calculate the similarity between users according the potential latent vectors.Finally,the calculated similarity is introduced into the random walk model to get the Top-k friend recommendation list.This method not only constructs the user model more comprehensively,but also considers the user interest and the user's social relation.Therefore,the accuracy of the recommended result is ensured.And the experimental results show that this method has a better performance than the existing friend recommendation methods.(2)This paper proposes a friend recommendation method combined the relationship of users with the method of learning to rank in social network.In this method,the relationship between users is calculated based on the number of mutual friends and the network structure of three-degree connection.Then,the relationship between users and the method of learning to rank is integrated into the latent semantic model to explore the potential preference characteristics of users.Finally,we construct the user sharing preference feature vector through the latent features of the related items shared by the user,and calculate the similarity of the feature vectors shared between the target user and others.And then we can get the target user the most similar orderly recommended list.On the one hand,this method is more comprehensive considering the influence of the relationship between the users.On the other hand,it is considering the influence of user's preference order for the items.Combined all of this can be better mining the user preference characteristics.The experimental results show that this method has achieved better results.