Collaborative filtering based on user attributes and user ratings for restaurant recommendation

Abstract

Online recommendation service had brought economic benefits for traditional catering industry. Aimed at the status quo, user-based collaborative filtering (UCF) algorithm was applied to restaurant recommendations in this paper. However, users' preference about restaurant was affected by many factors, leading traditional UCF algorithm precision was low. In order to solve this problem, three improvement were proposed. Firstly, mean score was enhanced to the calculation of similarity. Secondly, the number of common items between two users was utilized to affect the credibility of similarity, so modification factor was added to weaken the pseudo similar error. Finally, the real personal information online users registered were used to calculate the similarity based on users' attributes. The experimental results show that the modified algorithm (AdvancedCF) can improve the accuracy of the similarity calculation and provide users with more accurate restaurant recommendations.