Research And Implementation Of Course Recommendation System Based On Bilateral And Cross-domain

In recent years,with the rapid development of the internet,people have been able to come into contact with more new things and learn more knowledge,leading to the emergence of online learning platforms.The arrival of the epidemic era has brought the development of online education platforms to a higher level.Students can learn online through online courses,and a large number of users join the platform.A large number of courses are shared on the learning platform,bringing great convenience to users,However,it also generates many problems such as user cold start and inaccurate recommendations.The common unilateral cross domain recommendation only uses information from users or projects for relevant recommendations.The recommendation effect is good in areas that users have already learned,but there is still a problem of data sparsity in areas that users have not learned.Bilateral cross domain is a good solution to this problem,and the feature information obtained by unilateral recommendation is less diverse compared to bilateral cross domain.Therefore,a bilateral cross domain recommendation method based on reinforcement learning and graph neural network is proposed for cross domain recommendation.The user auxiliary domain and project auxiliary domain are set,and the graph neural network is used to extract user features for the user auxiliary domain.The graph neural network is used to extract project features for the project auxiliary domain,which is transferred to the target domain to achieve feature diversity and improve system recommendation accuracy.At the same time,in order to achieve long-term recommendations,a recurrent neural network is used for recommendation,and the next recommendation list is retrained based on user feedback on the items in the recommendation list.Simultaneously using reinforcement learning to train learning parameters for maximum long-term benefits.Improve the accuracy of the recommendation system while recommending.The model uses HR,Recall,and NDCG as evaluation indicators,and performs better than related single domain recommendations and bilateral cross domain recommendations in the dataset used.After fully studying existing theories,this system found that data sparsity and other issues can lead to inaccurate recommendations on the course recommendation platform.This study proposes corresponding solutions to solve these problems.The course recommendation platform is divided into two parts,namely the web client part and the backend server part.The overall structure is C/S,and Vue development technology and Spring Boot framework are used,The system development was completed using Java language and My Sql database.The client has implemented functions such as course recommendation,course comments,course Q&A,and personal information settings.In the server section,it mainly implements other functions such as administrator management of course resources and management of orders generated by front-end users purchasing courses.The system has passed the test and has high performance.