Research On Recommendation Algorithm Based On User Trust And Preference Information

Social recommendation algorithms usually integrate trust information in social networks in recommendation systems and then model social influence.Finally,users' interests are affected by their own preferences and trust friends' preferences,which effectively alleviates the problem of data sparsity and cold start in traditional recommendation algorithms.However,most of the previous work is to fuse trust information in matrix decomposition,which lacks the interpretability of recommendation and is too simple to measure trust.In addition,because deep learning technology has the advantage of extracting users and project features from other data sources,it can be applied to social recommendation algorithms to learn trust information.The key to social recommendation algorithms is how to quantify trust information and to combine trust information with the user's historical behavior preference information.The main contents of this paper are as follows:(1)In the task of rating prediction,in order to solve the simple way of measuring trust and the low interpretability of matrix decomposition algorithm problems,a Trust Regularization Euclidean Embedding(TREE)algorithm by fusing trust information was proposed.Firstly,the Euclidean embedding model was employed to embed the user and project in the unified low-dimensional space.Secondly,to measure the trust information,both the project participation degree and user common scoring factor were brought into the user similarity calculation formula.Finally,a regularization term of social trust relationship was added to the Euclidean embedding model,and trust users with different preferences were used to constrain the location vectors of users and generated the recommendation results.In the experiments,the proposed TREE algorithm was compared with the Probabilistic Matrix Factorization(PMF),Social Regularization(So Reg),Social Matrix Factorization(Social MF)and Recommend with Social Trust Ensemble(RSTE)algorithms.When dimensions are 5 and10,TREE algorithm has the Root Mean Squared Error(RMSE)decreased by 1.60% and5.03% respectively compared with the optimal RSTE algorithm on the dataset Filmtrust.While on the dataset Epinions,the RMSE of the TREE algorithm was respectively 1.12% and1.29% lower than that of the optimal Social MF algorithm.Experimental results show that the TREE model can further alleviate the problems of data sparsity and cold start,and improve the accuracy of score prediction.(2)In the top-k recommendation task,in order to use deep learning technology to measure trust and model social influence differently,an algorithm based graph neural network for the social recommendation was proposed,which integrated user trust and preferenceinformation.The algorithm first combined the user social vector extracted by the Autoencoder and the user potential feature vector in the user-item interaction matrix and used the attention mechanism to measure the trust between users.Then,on the basis of using a graph convolution network to model users' social diffusion,considering the influence of different trust users on the target users' preferences and the trust degree between users,the accurate user feature vector was obtained.Finally,the Multilayer Perceptron was used to model the complex interaction between users and items and generate the recommendation results.Experimental results show that the proposed algorithm has a certain improvement in recommendation performance compared with the related algorithms.