Recommendation Algorithm Integrating Social Network And Knowledge Graph

In view of the data sparsity and cold start problems of traditional recommender systems,many scholars begin to study how to improve the performance of recommender systems by using the social information of users and the semantic information of the items themselves.In recent years,the rapid development of social networks has brought abundant additional information to recommendation technology,and the social relationship and trust relationship between users have become an important factor affecting the performance of recommendation system.Graph convolutional neural network has been widely used because of its ability to vectorize graph data such as social network.In addition,with the continuous development of knowledge graph technology,more and more recommendation systems use the rich item attribute and semantic information contained in knowledge graph to improve the accuracy of recommendation algorithm.The learning method of knowledge graph representation can represent the semantic information of objects quantitatively and then facilitate various work,such as using vector matrix to calculate the similarity between objects.Based on this,this paper will focus on the research of recommendation algorithms based on knowledge graph and social network,and introduce the potential characteristics of similar users and similar items into the recommendation to promote the accuracy.The main contents are as follows:1.For the sake of figuring out the problem of low accuracy of recommendation,a social recommendation with social network user similarity is proposed based on comprehensive consideration of the influencing factors of social network user similarity such as social network subgraph topology,user trust and user rating similarity.Algorithm in the framework of traditional matrix decomposition,the first to use figure convolution neural network to users of social networks learning get involve social network graph topology structure and the connection relations of user characteristics,potential and the social relationship is used to calculate the user social trust,then the score data is used to calculate the user rating similarity,the integrated use of user characteristics,potential users trust and rating similarity computing social network user similarity and blend in the matrix decomposition,to predict the user to predict score value of the project.Experiments on public datasets reveal that the model promotes recommendation accuracy.2.learning method of knowledge map based on the translation of the representation,the semantic information of the entity of knowledge map items embedded into the dense lower dimensional vector space for vectorization,said information semantic relation for the items in the recommendation system,and then combined with user interaction-items information to strengthen the similarity between objects,this paper proposes a social recommendation based on the knowledge graph is to improve the performance of the recommendation.Based on the social recommendation that integrates the similarity of social network users,this algorithm introduces the information of similar item sets to improve the recommendation performance,and the MAE and RMSE values of Film Trust data sets are increased by 1.3%and 1.6% respectively.