Research On University Library Recommendation Algorithm Based On Collaborative Filtering And Markov Process

Library is a key building site in colleges and universities,every year has introduced a large number of books.How to help readers find their favorite choices in a vast smoky book is an urgent problem.Manual recommendation and traditional book search function all rely on past experience too much,and these experiences can not be stored,so that more people can benefit from it.Therefore,the establishment of personalized recommendation system in university libraries is of great significance.At present,there are not mature personalized recommendation systems in domestic academic libraries.The main reasons lie in the following points.First of all,book information is not enough,lacking content,abstract and other text information.It can not use content based recommendation based on topic model or text analysis;Secondly,the university library has no scoring system,which can not get the readers' preference,so it is difficult to use collaborative filtering recommendation;Finally,considering the identity of students,university readers should have special recommendation requirements,but the recommended algorithm is not yet implemented.In order to solve these problems,the following contributions have been made in this paper:(1)Propose a collaborative filtering recommendation algorithm in the model and the type based on factor.The algorithm builds an interest model through reader's lending behavior data to approximate the reader's score,so as to solve the problem of lack of scoring system in university library,and on this basis,we use the weight factor of Chinese graph to solve the problem of sparse lending relationship.(2)Analysis of university students' recommendation requirements,propose the single point recommendation problem.The problem is that students should have knowledge structureconnection and acceptance between books borrowed from different periods,rather than solitary points.(3)According to the single point recommendation problem,propose a professional book tree recommendation algorithm based on Markov process.The algorithm obtains the transfer probability matrix by statistics of the association degree between different term books,and obtains the books to be recommended every semester through the initial type vector of the first term multiplied by the transfer probability matrix.(4)Using the two recommended algorithms proposed in this paper and combining the MapReduce computing framework in the Hadoop distributed system,an university library recommendation system based on the historical data of the existing borrowing system is designed and implemented.Through the experimental proof and the example analysis,the two kinds of University book recommendation algorithm proposed in this paper have a certain recommendation effect,and have the practical value.It provides a new way of thinking for the research of college books recommendation.