Research On Book Recommendation Method Based On Implicit Feedback Of User Behavior In Large-scale

The recommendation system based on collaborative filtering(CF) algorithm makes it possible to recommend user personalized books, which eases reading and enhances reading efficiency. But many current e-reading systems for book recommendations lack of users’ rating data and even if there was one,the rating data was very likely to be inaccurate or sparse,which hinders the application of CF.To address above problem, we utilize the big data of massive users’ reading behavior.Firstly, the data is preprocessed by Hadoop distributed platform and the implicit reading behavior is analysed. The data derived from preprocessing stage is used for the purpose of further statistical analysisand modeling. Next, to solve the rating-deficient problem, we come up with a method where the implicit-reading-behavior is used. The experiments proved that this method is more accurate than traditional methods.The main work and results are summarized as follows,(1) To deal with the efficiency issues on large-scale data, we used MapReduce data processing model on Hadoop distributed platform. Firstly, the large-scale data is filtered and cleaned to obtain effective implicit-behavior data. Then the preprocessed data get further analyzed. The method of processing large-scale data improves working efficiency and enhances data accuracy.(2) To solve the problem of inaccuracy or sparseness of rating data and improve recommendation accuracy, we propose the T-F Model. It converts valuable implicit-behavior data to rating data, fills the User-Book rating array and realizes the accurate collaborative filtering recommendation for books.(3) To evaluate the effectiveness of recommending method based on large scale implicit reading behavior data, we divide the data set into training data set and test data set. The training data set is used to predict the rating and generate recommendation. We compare our method with user-based CF, item-based CF and traditional rating recommendation algorithm,and the experiment results show that our method outperforms others in terms of the accuracy.