A Social Network-Aware Top-A/ Recommender System using GPU

摘要

A book recommender system is very useful for a digital library. Good book recommender systems can effectively help users find interesting and relevant books from the massive resources, by providing individual recommendation book list for each end-user. By now, a variety of collaborative filtering algorithms have been invented, which are the cores of most recommender systems. However, because of the explosion of information, especially in the Internet, the improvement of the efficiency of the collaborative filling (CF) algorithm becomes more and more important. In this paper, we first propose a parallel Top-;V recommendation algorithm in CUDA (Compute Unified Device Architecture) which combines the collaborative filtering and trust-based approach to deal with the cold-start user problem. Then based on this algorithm, we present a parallel book recommender system on a GPU (Graphics Processor unit) for CADAL digital library platform. Our experimental results show our algorithm is very efficient to process the large-scale datasets with good accuracy, and we report the impact of different values of parameters on the recommendation performance.