| Libraries are one of the most important academic places of universities and colleges.Millions of books provide rich resources for teachers and students to study and research.However,in the face of massive collections of books,simple book retrieval is difficult to provide personalized services in a targeted manner.It’s hard to help readers find their favorite books effectively.As a result,the phenomenon that the collection of books is more important than use in college libraries is increasingly prominent.The technology of personalized recommendation can explore the potential needs of readers and actively recommend books for their preference.This technology can increase the overall borrowing rate of university libraries.Therefore,studying the personalized recommendation algorithm of university libraries has important theoretical and application value.This research focuses on some problems excited in traditional collaborative filtering recommendation algorithm,such as migration of reader’ s preference and sparseness of borrowing record.Based on latent factor model(LFM),combined with time decay function and K-means clustering algorithm,using the historical borrowing records as the main data,an in-depth study of college book recommendation algorithm is carried out,two improved algorithms are proposed.The core research work of the thesis mainly includes the following two parts:1.Aiming at the problem of readers’borrowing preference transfer,the LFM recommendation algorithm fused with time information(TIL)is proposed.First,build a comprehensive preference model based on the borrowing records and effectively transform the records into a reader-book comprehensive preference matrix;Second,based on the matrix,the Newton cooling formula is used as the time decay function to explore the migration of readers’ borrowing preferences.In this way,a reader-book revision preference model that integrates time information can be constructed to predict and recommend preferred books for readers.The results show that the TIL algorithm has better quality in book recommendation.2.Aiming at the problem of data sparseness,the LFM recommendation algorithm fused with time information and clustering optimization(CTIL)is proposed.According to the rules of Chinese Library Classification,combined with K-means clustering algorithm to improve,the constructed coarse-grained book category preference matrix is clustered for readers,and readers with similar book borrowing preferences are classified into the same cluster.This can effectively reduce the sparsity of borrowing records and optimize the data structure of the preference matrix.Then,based on the matrix of the same cluster,the TIL recommendation algorithm is used to predict the preference score and recommend books for readers.The experimental results show that the algorithm further improves the book recommendation effect than the TIL recommendation algorithm.Figure[18]table[13]reference[43]... |
