| With the rapid development of the Internet,the amount of information in the network has become very large and it is difficult for users to find the information they are interested in.The recommender system can study the users' preference and predict users' behavior and then recommend the information or items that the user may be interested in,which greatly solves the problem of information overload.The collaborative filtering algorithm has been widely used in recommender system because of its simplicity,high efficiency and strong interpretability.However,the traditional collaborative filtering algorithm has little research on the interaction pattern between users and the historical behavior of users themselves,limiting the performance of the recommender system.This paper proposes solutions to these problems and combines the social relation between users into the recommender algorithm to further improve the accuracy of the recommender system.The research work and innovations of this paper are as follows:(1)This paper analyzes the propagation form of users' influence in the network and proposes a recommender system based on information propagation model and implicit feedback.In addition to explicit effects such as ratings,there is an implicit influence that spreads among users who rated the same items.The paper studies the specific form of this propagation and introduces it into the recommender system.At the same time,the explicit trust relationship is integrated into the matrix fraction framework of the recommender system,which further improves the quality of the prediction.A series of experiments in two real data sets show that this model has a significant improvement over other state-of-art recommender model.(2)This paper extracts the nonlinear relationship between users and items from a deep level,analyzes the influence form of users' historical behavior on current behavior,and proposes a recommendation system based on artificial neural network and attention mechanism.By using artificial neural network,the system can obtain more nonlinear factors to simulate the interaction between users and items.In addition,the paper creatively uses information propagation model to simulate the users' own historical behavior affected by other users in the network,so as to realize the attention mechanism and make recommendations with users' trust relationship.The result of experiments indicates that the predication quality of this system is better than other neural collaborative filtering models. |
