Research On Electronic Book Recommendation System Based On User Behavior

With the rapid development of mobile Internet,the number of mobile Internet users in China has reached 1.25 billion.There are about 378 million users of online literature,over 13 million original works of online literature and 500 billion yuan of online literature market value.But fewer than 50 percent of the 378 million users were involved in commenting and scoring books,and book recommendation relies only on user initiative scores,unable to gain a migration of their interests.Therefore,how to get users' interest in books is a matter of concern.Based on the data of user's reading behavior and user's comment behavior,this thesis carries on the related research in the book recommendation system,used the reading behavior data to solves the problem of not participating in the scoring of the book review user,and optimizes the problem of the interest transfer of all the users.The main research work of the thesis:(1)By studying and analyzing the existing technologies and theories of user behavior analysis,emotion analysis and recommendation system,the feasibility of using user behavior analysis and emotion analysis technology in recommendation system is verified.(2)By analyzing and comparing the different behavior data of all users,reading time and reading speed,which are most influenced by the users' love degree of books,are obtained.And set up a score transformation model for the two behaviors--Time model and speed model.Through a series of analysis,this thesis combines the two scoring models linearly to get the ideal user-book scoring model,real-time long-speed model.(3)For the users participating in the book review,the Time-Speed model is directly used to recommend,although it can solve the problem of user interest transfer,but the recommendation accuracy is not as good as using the comment score to recommend.Therefore,this thesis uses the method of emotional analysis to convert the data amount of book review into emotional score and optimizes the Time-Speed model by using emotional score,which is a linear combination of the two.Finally,the score obtained from the optimized model is put into the recommendation system using collaborative filtering method.The experimental results show that the time-speed model of emotion score optimization proposed in this thesis is used in book recommendation system to improve the accuracy of book recommendation.Using the improved model as the input data of the recommendation system,the accuracy rate of recommendation for users who do not participate in book reviews is increased by 10%,and the accuracy rate for users who participate in book reviews is increased by 7.5%.