Research On Popular Books Ranking For Recommendation Method Based On Cluster Analysis

With the development and application of Computer Networks Technology,people inclined to buy books and publications in the online bookstore,such as Dangdang or Amazon,etc.The online bookstore provides fast and convenient way of shopping,further,according to user preferences or top list,books and publications could be recommended to the customers with friendly interface.Coupled with this tendency,for the online bookstore,to rank and recommend popular books personally and accurately to the customers according to the book database entries and user preferences could help customers find the latest books and publications conveniently and expand book sales to increase economic efficiency.On the foundation of clustering analysis method of data mining,this paper presents a recommendation method of personalized popular books based on the user evaluation habits.This paper classifies books and publications in the online bookstore according to factors of varied consumption habits and preferences for books.It implements personalized ranking by mapping various books into sorts of people.In the process of recommending and personalized ranking,the following methods are adopted based on traditional collaborative algorithm:1.Use the TF-IDF formula to extract the key words of books and publications,transform the key words into quantifiable feature vectors via the narration of the book,and set up customers’ evaluation matrix of the key words simultaneously.It provides the quantitative basis for the follow-up recommendation and similarity calculation.2.Introduce clustering analysis method of data mining into customer group analysis.It realizes predict marking to various books by sorts of customers.3.Propose a personalized recommendation method.By combining the purchase history of the customers,this paper carries out a similarity comparison on feature vectors and customer’s preference vectors,thus realizes the personalized recommendation.To determine the method presented in this paper,we conduct an experiment and analysis the outcome on the data of an online bookstore web system,and make a comparison between traditional KNN method and the method presented in this paper.It is concluded that the method presented in this paper could substantially improve the accuracy and recall rate of recommendation with limited recommendation quantity.