Research And Implementation Of Recommendation Algorithms Based On Social Network

With the development of Internet technology,massive amounts of data are generated every day,and people have also transitioned from the era of lack of information to the era of information overload.As a bridge between users and data providers,the recommendation system alleviates the dilemma of data overload.It can dig out users' interests from users'behavior patterns and other auxiliary information.In addition,with the development of social media,users have generated rich interactions through social platforms such as Weibo and WeChat,such as following,forwarding,and trusting,thus forming a huge social network between users.At the same time,users' preferences for information are easily affected by their friends on the social network.Introducing social network information into the recommender system can alleviate the problem of data sparsity and provide more accurate recommendation results.Therefore,how to fully integrate social networks and the interaction information between users and items and other auxiliary information so as to more accurately dig out the interests and preferences of users has become a very important issue.Based on this background,this research,based on graph deep learning technology,explores recommendation algorithms based on social networks.It mainly includes the following three aspects:First,a recommendation algorithm based on social networks that integrates temporal information is proposed.The algorithm models users and items from the perspective of networks.Firstly,users and objects are modeled through a graph model that integrates real time information.Then utilize the self-attention mechanism to model the sequential information and the real time interval information among items to characterize the user's dynamic preferences.Finally,the graph attention mechanism is used to model the user's social relationship.Finally,experiments are conducted to verify the effectiveness and accuracy of the algorithm.Second,a social recommendation algorithm that integrates knowledge graph information is proposed.Firstly,extract the entity-item subgraph from the knowledge graph for the target user and item,and propagate the embeddings of entities and items on the subgraph based on the graph neural network model.Secondly,by aggregating the embeddings of items in the user's historical purchase behavior,the latent representation of the user's preference is obtained.Then,the user's social relationship is used to aggregate the interest representations of his neighbors.Finally,the above representations are combined to jointly solve the problem of attribute inference and item recommendation.Experiments are conducted to verify the effectiveness of the algorithm.Third,a prototype recommendation algorithm system was designed and implemented.The system implements the recommendation-related modules and performs functional tests to verify the effectiveness of the platform and related algorithms.