Research On Movie Recommendation Algorithm Based On Social Network

With the rapid development of big data technology,the problem of "information overload" is becoming more and more serious.Especially in the field of movies,such problems are more prominent.Recommendation system,as one of the important means to deal with the problem of information overload,has attracted the attention of many scientific researchers.On the other hand,social network services are booming,and more and more movie recommendation platforms incorporate social attributes.In recent years,due to the powerful graph representation capabilities of graph neural networks,the recommendation system based on social networks has been given more opportunities for improvement.Therefore,this topic will use graph neural network technology to solve the task of movie recommendation based on social networks,and conduct research from two aspects of static social scenes and dynamic social scenes.The main work is as follows:(1)Aiming at the movie recommendation task in the static social scene,this topic studies a graph neural network social recommendation model CI-GNNSR based on coupled influence.First,in order to alleviate the problem of data sparseness,the user’s historical rating information and second-degree social information were mined,and a collaborative friend recommendation circle with similar tastes was constructed,and it was introduced into the learning modeling of the feature representation of users and movies.Secondly,in order to distinguish the correlation factors between different user’s friends,the attention-based memory network is used to learn the expression of interest and influence between users and collaborative friends.Finally,the Graph Rec model is improved by digging deeply into the influence of collaborative friends’ interests and the coupling influence of the movie to be tested on users.This model optimizes the user’s social influence from multiple angles.The experiment is verified on the Douban movie data set.Through comparative analysis,the algorithm studied in this topic is superior to other comparative algorithms on the two evaluation indicators of MAE and RMSE.(2)Aiming at the movie recommendation task in dynamic social scenes,this topic studies a graph neural network social recommendation model FSDFR-GNNSR that integrates static and dynamic feature relationships.First,use graph embedding algorithm to extract static features of users and movies.Secondly,improve the modeling method of the user’s dynamic characteristics,and use its static characteristics as part of the GRU network input,so that the model can consider the static characteristics while modeling the user’s dynamic behavior.Finally,the graph attention network is used to model the dynamic influence of friends,and the time complexity of the algorithm is reduced by simplifying the update strategy of the second-order neighbor nodes,and at the same time,the graph pooling operation is used to improve the generalization ability of the algorithm.The model comprehensively considers the user’s static and dynamic feature information,and enriches the information of friends.The experiment selects the movie category in the Epinions dataset for verification.The results show that the algorithm studied in this topic is compared with other comparison algorithms in HR value and NDCG The value has been improved.(3)Based on the theoretical research of this subject,a prototype system for movie recommendation based on social networks was designed and developed.The system introduces the idea of modular development and object-oriented programming,and completes the design and implementation of business processes and system interfaces.At the same time,the recommendation algorithm studied in this subject is tested and analyzed,which proves the effectiveness of the algorithm.