An Approach for Top-k Recommendation Based on Trust Information
Xiong Li-rong, Ling-yan Wang, Yu-zhu Huang · 2017
Personalized recommendations can effectively solve the data explosion problem in network. Most existing works utilize rating information to reduce the score prediction error, e.g. MAE; however, users prefer a list of top-k items and minimizing MAE does not always result in better top-k item lists. Meanwhile, because of data sparse problem, social connections among users play an increasingly important role in top-k recommendation system. So, this paper presents a new trust-based top-k recommendation algorithm, called BTRank. It integrates rating and trust information to construct the rating sorting model, which effectively improves the quality of the top-k item list of all users. A series of experiments on real-world datasets prove the effectiveness of our algorithm.