Research And Implementation Of E-learning Group Recommendation System Based On Graph Aggregation

At the beginning of 2020,in order to prevent the spread of the COVID-19 epidemic,the Ministry of Education issued the notice that delay opening the school of spring semester.However,the countermeasures of "classes suspended but learning continues" by the Ministry of education have further popularized E-Learning.E-Learning be just unfolding,which also makes K12 online education industry develop rapidly.With the development of Group Learning theory and the grand success of offline practice,the form of Group Learning has been promoted continuously.The existing E-Learning systems are all for individual learners,few of them introduce Group Learning into E-Learning systems,and there is not group recommendation function.Secondly,the current group recommendation system can’t provide group recommendations for group members fairly and satisfactorily.In addition,the subject types of learning resources are distributed unevenly,with more resources for cultural courses and less resources for artistic interests;Finally,the typical problems of E-Learning which made the students become less interested and get less efficient have become increasingly prominent.Through in-depth study of Online Education theory and Group Learning theory,combined with professional knowledge and big data related technologies,this thesis introduces group learning into E-Learning system and provides recommendation function for learning groups.This system adopts the design pattern of front-end and back-end separation,the front-end uses the lightweight framework Vue,the back-end uses the Java’s "out-of-the-box" framework Spring Boot,the persistence layer uses the simple and flexible framework My Batis,and the algorithms of the key technologies were implemented in Python.The key technology part of the system is proposing a general group recommendation model.About implementing group recommendation,we first fill the user-item matrix by Bayesian personalized ranking which based on matrix factorization,and then get the prediction score matrix of user-item.Then the similarity of users is calculated by Pearson correlation coefficient for group users.With the user-item prediction matrix and assigned group,we propose a method of expressing user preference by directed graph,and propose Graph Aggregation algorithm to aggregate the preference graph of group members,so as to obtain the result of group recommendation.Finally,the experiment shows that graph aggregation algorithm is superior to common aggregation strategies in the performance of aggregation results.Finally,this thesis realizes an E-Learning system which can be used for student users to learn courses online,exercises online,study in groups online,and there is group recommendation function of learning resources.It enriches the E-Learning methods,solves the problems of online learning student,like "fuzzy demand" or "difficult to choose",and improves students’ interest and quality of E-learning.