CoRe: Exploiting the personalized influence of two-dimensional geographic coordinates for location recommendations

With the rapid growth of location-based social networks (LBSNs), location recommendations play an important role in shaping the life of individuals. Fortunately, a variety of community-contributed data, such as geographical information, social friendships and residence information, enable us to mine users' reality and infer their preferences on locations. In this paper, we propose an effective and efficient location recommendation framework called CoRe. CoRe achieves three key goals in this work. (1) We model a personalized check-in probability density over the two-dimensional geographic coordinates for each user. (2) We propose an efficient approximation approach to predict the probability of a user visiting a new location using her personalized check-in probability density. (3) We develop a new method to measure the similarity between users based on their social friendship and residence information, and then devise a fusion rule to integrate the geographical influence with the social influence so as to improve the user preference model on location recommendations. Finally, we conduct extensive experiments to evaluate the recommendation accuracy, recommendation efficiency and approximation error of CoRe using two large-scale real data sets collected from two popular LBSNs: Foursquare and Gowalla. Experimental results show that CoRe achieves significantly superior performance compared to other state-of-the-art geo-social recommendation techniques.