Research On Personalized Recommendation Of University Library Based On Cluster Analysis

The thesis aims to realize the research on universal method of personalized recommendation for university libraries based on clustering algorithms.By using several basic recommen-dation algorithms,design a personalized recommendation strategy which is suitable for uni-versity libraries.Book borrowed history records in 2017 and undergraduates(2014-2017)'personal information are used as instance data to perform data mining to verify the feasibility of the personalized recommendation strategy designed previously.Firstly,establishes a Reader-BookClass preference model,and performs cluster analysis on target readers based on the preference model.In the process of cluster analysis,the basic K-Means clustering algorithm is optimized,and then the optimized K-Means algorithm is used to implement reader clustering.Secondly,collaborative filtering recommendation algo-rithm and content-based recommendation algorithm are combined to achieve the purpose of realizing the advantages of both recommendation algorithms simultaneously.Use collabora-tive filtering and content-based combined recommendation(interest-based recommendation)strategy to generate a personalized recommendation booklist for each target reader.In addi-tion,the thesis uses non-personalized recommendation strategies for new readers who don't have enough borrowing records.The non-personalized recommendation strategy is based on the faculty and major of each target reader.It also generates a recommendation booklist which consists of popular books in faculty and major.Through the design of these two as-pects of recommendation strategies,personalized recommendation for university libraries is finally achieved.In the research process,readers' borrowing preference and borrowing char-acteristics could help Central China Normal University Library to provide higher quality and humanized services to students.The main work of the thesis is as follows.Established reader preference model based on books classification number in CLC(Chinese Library classification).Give a detailed introduction of CLC.The basic k-means algorithm has been optimized,and the experimental comparative analysis is performed using the in-stance data.Detailed description of different recommendation algorithms and strategies.Reader cluster-ing is based on the readers' preference model,and then,collaborative filtering and content-based combined recommendation strategy is used to generate a personalized recommenda-tion booklist.In addition,for new readers,non-personalized recommendations are used,that is,based on the faculty and major where the readers are in.The thesis optimized basic K-Means algorithm by using rotation algorithm in the process of selecting center points of initial clusters.Index of DB and RMSSTD are used to evaluate the performance of the optimized K-Means algorithm.The experiment shows that the personalized recommendation of the university library de-signed in this thesis is feasible.Meanwhile,there also are imperfections in the personalized recommendation strategy.