Research And Implementation On Knowledge Concept-driven Explainable Course Recommendation In MOOC

With the rapid development of the Internet and the growing demand for education,a new online learning model of massive open online course emerged in 2012.With the rapid development of large-scale online course platforms,anyone can upload courses to the platform.The number of courses continues to increase and there is a large amount of redundancy in teaching content.There is a problem of information overload similar to ecommerce platforms;at the same time,because the online open platform lacks a clear learning structure to guide users to develop learning strategies,it is difficult for users to choose a course that suits them from the massive amount of data.Therefore,this article aims to use the recommendation system to analyze the user’s historical behavior,effectively mine the user’s preference expression,and realize the personalized course recommendation for users in the massive open online course application scenario.However,the research of recommender systems faces many challenges.On the one hand,in the field of education,the scale of users on the platform is very large,but their interaction data is very scarce,and the data is facing the problem of sparseness.On the other hand,most of the existing studies model users as static representations,which are not suitable for the application scenarios of course recommendation,because the learning behavior of users is a dynamic and continuous process,and static modeling cannot well represent the learning growth of users.Finally,the existing recommendation system technology mostly pays attention to the accuracy of the recommendation results and ignores the interpretability.The interpretability can not only enhance user trust,but also provide improvement directions for model designers.In response to these problems,this article proposes the following three methods:(1)Aiming at the problem of data sparseness,this paper adopts the strategy of fusing knowledge concept information into recommendation system.To this end,a new data set—Java online learning data set is constructed,and a knowledge concept map is built to introduce the recommendation system,so as to introduce more effective information for the recommendation system.At the same time,in order to reduce the introduction of noise in the knowledge graph,we introduce a meta-path strategy to extract only the information that meets the needs in the knowledge graph to participate in the recommendation process,thereby improving the recommendation performance.(2)Aiming at the problem that traditional static modeling is not suitable for course recommendation,this paper adopts time series modeling method,adopts gate recurrent unit modeling in the time dimension,and uses the user’s historical interaction data and user representation as the input of the gate recurrent unit,and the output is the probability of a user interacting with each course at a moment.(3)In the research of the existing knowledge graph fusion recommendation system,the path-based method can bring interpretability to the recommendation result.This paper uses the connectivity of the knowledge graph to connect users and courses through predefined meta-paths.At the same time,it learns different weights for different paths and distinguishes the contribution of each path to the recommendation results.Making the recommended results interpretable.Finally,I use the Java online learning data set to compare the traditional recommendation model 、 sequence recommendation model and the existing recommendation model fused with the knowledge graph.The results show that the model proposed in this paper is optimized in terms of recommendation performance and interpretability,and experiments verify the effectiveness of the proposed model.The research of personalized recommendation system not only has important guiding significance for users to obtain effective information,but also efficient recommendations can also enhance the market value of online services,and even have important implications for social development and national security.The research on the related theories and methods of the recommendation system combines different disciplines and branches,and promotes the development of different disciplines.It is also a typical representative of the combination of industry,university and research,which is conducive to promoting industry and scientific research to cooperate with each other and embodying comprehensive advantages.