Research On Design And Implementation Of Book Recommender System

With the development of Information Technology and Internet, people are moving from the era of information insufficiency to information overload. To solve the problem of information overload, recommender systems emerge. Recommender systems can recommend potential interesting items to users from a large item space. Recommender systems have been deployed successfully in a wide range of domains. But there are few applications of recommender systems to higher school libraries. Most libraries don’t have the functionality of book recommendations. They don’t extract and analyze readers’ interests from historic borrowing records, and they can’t provide personalization services to readers.This paper researches applications of recommendation technologies in the domain of book recommendations, in the hope that libraries can provide better service to readers. The recommendation technologies implemented in this paper include personalization technology and non-personalization technology. Personalization technology can recommend personalized books to different readers given the borrowing records of different readers. Collaborative filtering is the most widely used personalized recommendation technology. The personalization technology adopted in this paper is collaborative filtering. But collaborative filtering algorithms face new user and new item cold start problem. In addition to personalized recommendation technology, this paper implements non-personalized technology too. Non-personalized technology can provide high quality and timely recommendations to readers through expert recommendations, new book recommendations and etc. At the same time, non-personalized technology can solve the new user and new book cold start problem.Selecting different parameters(similarity measures, number of neighbors considered, number of books recommended) while using the collaborative filtering algorithm will produce different performance. This paper will discuss the effects of different parameters in order to select optimal parameters and produce optimal performance.Finally, collaborative filtering algorithms are either based on users, or based on items. This paper implements both and will discuss their performance.