Research On Personalized Course Recommendation Methods Based On Hybrid Topic Modeling

With the rapid development of network technology,education data analysis and other technologies,network communication continues to develop into the web3.0 era,and the continuous updating of technology brings new challenges and requirements to education.The sharp increase in demand for teaching students in accordance with their aptitude and personalized learning.In order to meet the needs of users’ personalized course learning,effective course recommendation methods have become a current research hotspot.The existing MOOC platform course recommendation method usually uses the user’s historical learning records and obtains its dominant subject area to describe the user’s preference model,and then complete the recommendation.This recommendation method has a good recommendation effect in the recommendation of courses in leading subject areas.However,users often need to learn courses other than the dominant subject area.In this case,it is difficult to accurately recommend courses in other subject areas based on the user’s own learning preferences.Aiming at the problems in the recommendation of online learning courses,this article has conducted an in-depth study,fully considering the historical learning situations of different users visiting the same subject area,or the same user visiting different subject areas,and obtaining user preferences and subject area preferences from it,and proposes a personalized course recommendation method based on mixed topic modeling.The main contents are as follows:(1)Propose an offline preference model(PS-LDA)based on personal learning preference and subject area preference,aiming to model the preference in a unified way considering the two factors of user’s personal learning preference and subject area preference.Specifically,P-LDA can automatically learn user learning preferences from user history learning records by subjecting users to topic modeling,and solve the problem of data sparsity to a certain extent.In order to further alleviate the problem of data sparsity,S-LDA uses popular courses in the subject field(that is,through subject modeling of the learning records of all users who visit the subject field)to generate the subject field preference.In this way,the recommendation system can not only be used for the user’s course recommendation in user-led subject area,but also can be well recommended in other subject areas.(2)Propose an online recommendation strategy based on the ranking framework.Using offline modeling to query the user’s personal learning preference model(P-LDA)and access subject area preference model(S-LDA),and considering whether there are overlapping topics between the models,the online ranking framework is used to calculate the number of courses to be recommended Scoring,combined with the threshold-based quick sorting method,speeds up the top-k course item generation process,and completes the recommendation.(3)Experiments on the ed X and GCSE datasets show that,compared to models that only consider users’ personal learning preferences,the PS-LDA model can more comprehensively model user learning preferences,especially when users visit other subject areas with good accuracy.