Research On Personalized Recommendation And Group List Customization Algorithm For E-book

With the accelerated pace of life,e-book reading has become a way for people to study and entertain in their spare time.When faced with massive e-books,the recommended system and ranking list become the entrance to the reader to quickly find their own books interested in.However,in the case of ranking researches,most of them are based on the statistical data set,and thus lack the predictive recommendation to reflect the reading trend;In the e-book personalized recommendation,due to the lack of users' feedback data,so that the recommended accuracy is affected.Therefore,aims at these problems,this paper studies the two aspects of e-book: personalized recommendation and group-oriented list customization.In the personalized e-book recommended work,we mainly study the recommendation algorithm of e-book reading scene without users' active feedback behavior.Firstly,we propose a personalized recommendation algorithm based on users' implicit feedback.The algorithm not only evaluates the users' preference for books from the user's implicit feedback,such as reading time and reading frequency,but also takes into account the difference in the reading speed between the reading users.And then convert users' implicit feedback to rating score to complement the user-book rating matrix.Finally,the recommendation algorithm(CF or SVD)based on the rating matrix is used for personalized recommendation.Through the comparison experiment,it is proved that this algorithm proposed a certain improvement on the accuracy rate of electronic books recommended in each index.In terms of group ranking list recommendation,we present a Reading quantity-driven e-book ranking list customization algorithm.Unlike the personalized recommendation,the purpose of the algorithm is to develop a list of popular books for group users,maximizing the amount of reading on list books.For the current e-book list can not reflect the future of the book hot situation,the goal of our algorithm is to develop a popular e-book list can reflect the future reading books.This algorithm is divided into two steps: Time-Decay Random Forest model(TDRF)and Behavior Based Re-Ranking model(BBRR).Firstly,using the TDRF model to select the small sample e-books which may be on the list from the massive e-books,and then the BBRR model is used to reorder the selected small samples according to the user behavior.Finnally,we get the final list.Through the comparison of real data,it is proved that the method proposed in this paper can increace the reading amount significantly than traditional method.