Research On Personalized Recommendation Of Online Learning Partners Based On Fine-grained Interests

With the rapid development of Internet technology,online learning is becoming more and more popular among users.In 2020,the “COVID-19” epidemic has caused a large number of users to choose online learning.And the number of users has increased day by day,which has brought a huge amount of user information.The number of users and traffic are the core competitiveness of the online learning community.Recommending learning partners to target users can eliminate loneliness in the learning process,and increase users' activity as well as dependency on the learning platform.Therefore,how to accurately dig out the user's learning interest and recommend learning partners becomes particularly important.At present,there have been a lot of related research work on learning partner recommendation,but it still faces three main challenges:(1)Existing public learning data sets have sparse social network and user's fine-grained learning interest cannot be accurately extracted,what resulting in the poor effect of learning partner recommendation;(2)Existing learning partner recommendation methods ignore time evolution of the learning interest sequence makes it hard to predict the user's future learning interest,which leads to a decrease in the accuracy of learning partner recommendation;(3)Lack of an accurate learning partner recommendation method.In response to the above challenges,this paper proposes a learning partner recommendation model based on fine-grained interests.The main contributions are as follows:(1)Crawling online learning data sets and extracting fine-grained interest tags.We use web crawler technology to capture two learning data sets of China University's MOOC website and Douban Book,and propose a method for extracting users' fine-grained interests.In addition,we extract keywords from introduction texts to construct a tagging system,and use clustering algorithms to extract feature tags that can represent users' fine-grained interest.A large number of offline experiments have proved that fine-grained interest tags can effectively improve the accuracy of learning partner recommendations.(2)Modeling time evolution of user interest sequence.We analyze the features of the user's learning interest sequence,and model time evolution of the interest sequence by adding a time-aware controller and a semantic-aware controller to predict the user's interest tag.Experimental results show that the modeling time evolution of interest sequence can predict users' interest tag more accurately,and reduce the sparsity of user-item rating matrix.(3)Proposing a learning partner recommendation model based on fine grained interests.We propose a personalized recommendation model for online learning partners based on the fine-grained interests of users and the evolution of interest sequences.In addition,we analyze the effect of users' social influence and interest domain on the learning partners recommendation.Experimental results show that the learning partner recommendation method proposed in this paper performs better than the existing methods.