Research And Implementation Of Book Recommendation System Based On GPU

Recent years, Digital Libraries has received a high degree of attention in many countries. It has rapidly become an important approach for public to access knowledge and information. However, with the rapid growth of the information in digital libraries, it becomes more and more difficult to find out useful information from the massive resources. The traditional way to access information is based on the queries of the user's information, which doesn't meet individual needs of users. Therefore, personalized recommendation system is becoming an integral part of the Digital Library. This paper studies and presents a recommended algorithm for books, which is used to help users find interesting and relevant books from the massive resources. Then based on this algorithm, we present a parallel book recommender system on GPU (Graphics Processor Unit) for CADAL digital library platform.The main contribution of this paper is as follows:(1) A hybrid top-N recommended algorithm has been proposed to combine the user-based collaborative filtering and trust-based approach to deal with the cold-start user problem. The experimental results show that our hybrid algorithm performs absolutely better than user-based collaborative filtering and trust-based algorithm, especially for cold-start users.(2) We implement the parallel version of our hybrid algorithm on GPU with the help of CUDA. The experimental results show that our parallel algorithm based on GPU performs higher efficiency compared with the algorithm based on CPU.(3) We built a CADAL Parallel Book Recommender System (CPBRS) on a GPU for CADAL digital library, based on the hybrid algorithm aforementioned and implicit ratings extracted from users' access logs. Then we integrate it into the new CADAL Digital Library Personalized Service Platform (iCADAL) and achieve better recommendation effectiveness.