Leveraging Social Relationship To Improve Personalized Recommendation

With the development of social media,online users can easily share their experiences on products or services with friends.The social recommendation system can be used to personalize and customize products that fit user needs.Existing social recommendation methods try to consider social information to improve recommendation performance,but they do not distinguish the impact of different social information and do not conduct an in-depth analysis of social information.In this article,the research will be carried out from the perspective of the pluralism of socialization relationship and the strength of socialization relationship.Considering the diversity of socialized relationships,the personalized recommendation method uses friend information and group information to expand the traditional Bayesian personalized recommendation model.First,construct three sets of partial order hypotheses and divide the feedback set.Then,based on the hypothesis of partial order relationship,a Bayesian personalized recommendation model is constructed which combines friend information and group information.A personalized recommendation method that considers the strength of socialized relationships,and uses the hierarchical attention mechanism in deep learning to learn the weight of the influence of friend information,group information,and tag information on user preferences.For different users,the bottom-level attention simulates the influence of different elements on the characteristic representation of information,and the top-level attention simulates the influence weight of different information on user preferences.The experimental results show that the two methods proposed in this paper can effectively improve the accuracy of recommendation and reduce the problem of sparsity.At the same time,both methods conduct in-depth analysis of socialized information,which can promote the development of socialized commerce.