Research On Book Personalization Recommendation

The pinnacle of this period is big data,and the rapid growth of numerous technologies has caused things to begin to combine intelligent aspects.The book publishing industry is preparing for the information age’s arrival.It is inevitable that a series of improvements will be made,coupled with the big data era In today’s world,it is very difficult for readers to choose from a good deal of books to read what they want.Therefore,it is great significant how to quickly and accurately recommend books for readers to the right books for them.However,users are like a drop in the bucket for book-type e-commerce platforms compared with many books,and the behavior data of each book corresponding to the user is even rarer,so fewer users and user behavior seriously affect the accuracy of book recommendation.When the new books and new users log on to the book e-commerce platform,the platform has not yet begun to record their behavior data and their similar neighbor users,so that the recommendation effect will not be too ideal.When there is a cold boot problem,Elements such as interests and new indicators are added to the recommendation system to improve the Precision of the recommendation.The problems of data sparsity and cold boot are unavoidable in personalized recommendation.Therefore,this thesis designs a collaborative filtering similarity algorithm with data tags for personalized book recommendation and integrates it into long-short term memory network based on self-attention mechanism.The cold boot recommendation algorithm of,The main work includes the following two aspects:（1）A collaborative filtering similarity algorithm for adding data labels.This algorithm aiming at the problem of data sparsity and long running time of personalized recommendation,this thesis focus on how to use the principles and techniques of collaborative filtering to reduce data sparsity and some negative effects caused by cold boot in book recommendation systems.Based on the tagging recommendation algorithm,in view of the cold boot recommendation problem of new books.The first is to filter the processed persistent data set through the tagging collaborative filtering algorithm,and store the processed data results in the Redis middleware.In this way,frequent access to the database for reading and writing is reduced,and the efficiency improved.Through experimental comparison,the accuracy of personalized recommendation is improved to a certain extent.（2）The recommendation algorithm of long-short term memory network based on self-attention mechanism is proposed for the storage bottleneck problem caused by the long-term storage of information,considering that the storage of non-keywords should be ignored,and the long term dependence between information should be reduced.We incorporate the self-attention mechanism into it.Through the establishment of the Self-Attention-LSTM model,the feature word extraction is performed on the existing result set again.Calculate the similar preferences of users and items and make recommendations by ranking.Experiments show that this algorithm is superior to the traditional recommendation algorithm in accuracy.The problem of recommendation accuracy of recommender systems is solved to different degrees.