Application Of Improved Collaborative Filtering Algorithm Application In Course Resource Recommendation System

Nowadays,information technology and education are rapidly integrating,and the familiar offline education model can hardly meet students’ demand for personalized learning resources.Especially during the epidemic period,online education has become an important supplement to traditional education,and this trend is also constantly promoting the development of online education industry.As a result,a variety of online course resources also flood into the online education platform,resulting in a serious "overload" of course information.During the learning process,users will obviously feel that the course resources on the platform are too complicated and disorderly,and it is not possible to give personalized guidance to the course selection according to the characteristics and needs of users.It is easy for users to get lost in the search for a large number of online course resources.The final learning efficiency is greatly reduced.In order to accurately depict the user portrait,collaborative filtering algorithm is proposed to recommend personalized course resources for users.However,in the practical application process,the algorithm has many problems that affect the performance of the recommendation system,such as poor expansion,cold start,Matthew effect,data sparse and so on.Therefore,this paper improved the traditional collaborative filtering algorithm on the basis of it,and then weighted the improved algorithm to get a mixed recommendation algorithm,and finally designed and implemented the course resource recommendation system.The main work of this paper is as follows:(1)Study the current status of course resource recommendation platform,read literature related to recommendation system and recommendation algorithm,study the research status and related theories of collaborative filtering algorithm and hybrid recommendation algorithm,and explore the improvement and verification of algorithms by some outstanding scholars.(2)A collaborative filtering recommendation algorithm based on user feature weight confidence is proposed.On the one hand,this algorithm alleviates the cold startup problem caused by the absence of historical behavior data of the traditional collaborative filtering algorithm when encountering new users;on the other hand,it also reduces the impact of poor recommendation effect caused by the sparsity of the user score matrix.(3)A collaborative filtering recommendation algorithm based on the popularity correction factor is proposed.This algorithm determines whether to correct the project by judging whether the popularity of the project exceeds the corresponding threshold.The principle is to reduce the proportion of popular projects in the algorithm calculation process,so as to recommend some projects that users are interested in more personalized.Then,appropriate values of the weight factors of the above two improved algorithms are determined and fused to obtain the final hybrid recommendation algorithm weighted by the improved algorithm.Through experimental comparison,it can be proved that the performance index of the improved algorithm in this paper is significantly improved compared with the traditional collaborative filtering algorithm and the algorithm improved by other scholars,and it has better score prediction effect.(4)Finally,the improved algorithm is applied to the course resource recommendation system,and the requirements of the system are analyzed according to different functional modules,and the database table is designed,and the development of the whole system is realized by using Spring,Vue,Axios and other technical stacks.