Research On Intelligent Construction Of Online Learning Community And Group Recommendation

Building an online learning community can improve learners’ online learning experience and enhance learners’ sense of satisfaction and sense of belonging.At present,most of the online learning community construction methods use a single construction strategy,such as the use of heterogeneous grouping or homogeneous grouping,a single construction strategy has some disadvantages.At the same time,there are a large number of learning resources in online education,but it is difficult for learners to quickly choose high-quality learning resources that meet their learning needs,which leads to the problem of information maze.Resource recommendation to members of the learning community can effectively alleviate the problem of information overload.The learning resource recommendation of the online learning community is different from the individual recommendation,it needs to consider the preferences of all members,so as to make the recommendation results meet the needs of all group members as much as possible.The traditional group recommendation adopts a predefined fusion strategy,which lacks the weight of dynamically adjusting the influence of members.In response to the above problems,this paper conducts research on the intelligent construction of online learning community and group recommendation methods.The main research contents and contributions are as follows:(1)The online learning community construction problem is abstracted as a multiobjective optimization problem,and the five main characteristics of learners are selected as the influencing factors for the construction of the learning community.A method of constructing learning community based multi-objective genetic algorithm is proposed.Ensure the "homogeneity between groups" and "heterogeneity within the group" of the learning community in terms of gender,learning style,and activity,as well as "homogeneity within the group" at the level of knowledge and hobbies.(2)We propose a group recommendation method based on hierarchical attention mechanism.We use a hierarchical attention mechanism for modeling,using a two-layer attention neural network,the first layer is used to capture the influence between members,and the second layer is used to capture the influence of members in decision-making,combined with the history of group interaction make group recommendations.This enables group members to have different contributions in group decision-making,and the weight of group members can be dynamically adjusted while taking into account the interaction between group members.Second,model the group’s topic preference,and learn from the group’s historical interaction data with the project to obtain the group’s topic preference.Finally,neural collaborative filtering is used for group recommendation.(3)Through comparative experiments,the effectiveness of the learning community construction method based on the non-dominated sorting genetic algorithm with elite strategy is verified.The learning community is constructed through this method,and the balance between the learning communities,the diversity and the fusion of the learning community are realized.It is verified that the recommendation effect of the group recommendation method based on the hierarchical attention mechanism is better than the effect of the comparison experiment.It shows that considering the interaction between group members can improve the recommendation results.