Research And Application Of Personalized Recommendation Method Of Books Based On Borrowing Record

With the increasing development of the publishing industry,the number of books and libraries in the library has been increasing in quantity and variety,and it is difficult for readers to find books of interest in massive books.At present,the library recommendation system of university library generally only rely on borrowing to recommend popular books,can not achieve personalized recommendation.Therefore,it is necessary to study the personalized recommendation method of university books.Based on the borrowing record of a university library in recent ten years,this paper designs a recommended method to realize personalized book recommendation.The method mainly includes two parts,the first part of the use of collaborative filtering algorithm to do the recommended results of the coarse recall.The second part constructs the reader preference model by extracting the feature of the borrowing record.The model is used to predict the books in the first part of the rough recall results,and the final recommendation result is generated according to the ranking.The principle of collaborative filtering algorithm is to find the similar users of the target user and recommend the target user according to the behavior of the similar user.The collaborative filtering algorithm relies on the user-project score matrix to calculate the user similarity.This paper takes the borrowing data of university library as the research background,based on the borrowing record to generate the reader-book scoring matrix to express the borrowing relationship between the reader and the book,fill the matrix with the number of times the reader borrows the book,express the reader's score on the book,Score normalization.Based on two kinds of collaborative filtering algorithms and two methods to calculate the similarity degree,the four algorithms are combined to make the comparison experiment.Mean Absolute Error(MAE)is used as the evaluation criterion,and the optimal algorithm combination is selected.The results of the recommendation included in the recommended users with different professional,different grades of readers borrowed books,to achieve a personalized book recommendation.In the second part of the method,the reader information,book information and borrow information to extract features.Select the borrowing records of all readers,sort by borrowing time,use the appropriate time window to construct the positive and negative sample set,use GBDT algorithm to train the data,construct the reader preference model,predict the first part of the rough recall results,Sort to produce the final recommendation.Finally,the author builds the book personalized recommendation system with the recommended method as the core,and the reader interacts with the recommendation system through the identity authentication to obtain the personalized recommendation result which accord with their preference.