Course Recommendation System Based On Graph Embedding Methods

In the Internet era,with the development of information technology,online education which holds the merits of flexibility and convenience has become a trend of future education.However,with the augmentation of users and resources on the online education platform,the phenomenon of so-called "resource explosion" will occur.Therefore,it is necessary to optimize the function of education platforms and enhance user experience by course recommendations.In this paper,the student online course selection data is taken as the research object,and the data is modelled into a studentcourse bipartite graph preparing for the research of course recommendation.To address the problems of insufficient node feature information extraction and insufficient relationship mining among multiple types of nodes in course recommendation,two course recommendation methods which combined graph embedding learning and classification methods are proposed,one is named a course recommendation method based on Node2 vec and bipartite graph projection and the other is named as a course recommendation method based on sampling fusion and Metapath2 vec.Besides,a couse recommendation platform based on the proposed graph embedding learning methods is established.The specific research contents of this paper are as follows:(1)A course recommendation method based on bipartite graph projection and Node2 vec.A student-course bipartite graph is constructed by extracting the relationship of students’ online course selection data.Then,the bipartite projection method is adopted to obtain the student projection and course projection.By implementing the Node2 vec on each projection,the feature of student and course is obtained separately.Based on the learned feature of two types node,a link prediction model based on logistic regression model is constructed for course recommendation.The proposed method is able to achieve sufficient feature extraction of student and course node by combining bipartie graph projection and graph embedding.Compared with the traditional methods,the proposed course recommendation method has a significant improvement in accuracy.(2)A course recommendation method based on sampling fusion and Metapath2 vec.Based on the feature of student-course bipartite graph with multiple types of nodes,Metapath2 vec is adopted to jointly mine the student and course information,thus student and course feature with a more complete information can be obtained.Besides,sampling fusion is used to enhance the algorithm’s ability to mine the local structural information of the data.The method is able to complete feature extraction of multiple types of nodes,which improves the accuracy of course recommendation prediction compared with that performed by single type of node feature extraction.(3)A course recommendation platform based on graph embedding learning.A course recommendation system platform is built based on the proposed graph embedding based course recommendation methods.According to the characteristics of different roles,the platform can customize the platform functions to improve users’ sense of platform use experience.The course recommendation function of the platform plays a certain auxiliary role in the realization of the high-quality utilization of teaching resources and the implementation of high-quality online education.