Research On The Construction Of Personalized Recommendation System Of College Book Resources Based On Multi-dimensional User Portrait

With the rapid development of computer technologies such as big data and artificial intelligence,the "smart +" service model is the development trend and direction of information services.The construction of smart libraries is an important area of smart services.It emphasizes user-centered and provides users with accurate information and knowledge services based on user needs.Nowadays,college libraries have collected a large amount of book resources,but most college libraries still provide traditional retrieval services.This one-way retrieval service is not efficient and accurate.With the rapid development of technologies such as user portraits and recommendation algorithms,university libraries can transform from passive services in which readers independently search for "people looking for books" to active services of "book looking for people" in which readers recommend books that meet their needs,enhancing the personalization of services,intellectual.Aiming at the problems of insufficient user portraits and poor algorithm performance in the current knowledge service field of university libraries,the paper studies multi-dimensional user portraits,book models and recommendation algorithms.According to the limited implicit feedback and explicit feedback,the users' reading needs are discovered,so as to improve the personalization and accuracy of library information services.The main tasks of the paper are as follows:(1)Combining the Chinese Library Classification Number and the Book-Label two methods,a book model is established,and a book similarity calculation method based on the book model is proposed.According to the characteristics of different user groups in colleges and universities,combined with the four dimensions of borrowing time,popularity,borrowing frequency,and borrowing characteristics,multi-dimensional user interest characteristics are proposed and user portraits are constructed to meet the precise service needs of university library users..(2)A Spark-based offline and real-time hybrid recommendation algorithm is proposed.In offline recommendation based on the book model,in order to effectively solve the cold start problem of the system,a content-based recommendation algorithm is proposed for new books and new users of the system,and the accuracy of user similarity calculation is improved through multi-dimensional user portraits and keywords extraction.Meanwhile,in order to effectively alleviate the problem of data sparseness,a collaborative filtering recommendation algorithm based on ALS is proposed for users with certain historical borrowing behavior data,and the matrix decomposition method is used to reduce the data missing rate and sparseness.In real-time recommendation,in response to the real-time requirements of the system and the changing interests and preferences of users,a real-time recommendation algorithm based on the book model is proposed.The "recommendation priority" is used to assign weights to candidate books,and the final recommendation results are sorted according to the weights.(3)Tested for offline and real-time performance with hybrid recommendation algorithm in this paper,by adjusting the parameters,the optimal model,experimental results show that the hybrid recommendation algorithm showed a more ideal in precision and recall;At the same time,compared to traditional stand-alone servers,the Spark-based hybrid recommendation algorithm is significantly more efficient in computing power.(4)Designed and implemented a personalized recommendation system for college book resources based on multi-dimensional user portraits,integrating core functional modules such as book recommendation,book tags,and personal space.The system can basically meet user needs in terms of recommendation accuracy and real-time performance in practical applications.