Abstract
Location-based services (LBSs) provide users with points of interest (POIs) based on a limited distance from the user current location. Recently, the increased number of POIs in LBSs often results in overwhelming place suggestions generated to users more than before. To facilitate a user's exploration, POI recommendation is demanded. Unlike traditional recommendations, POI recommendation has no explicit user rating and location awareness. We propose in this paper a novel method to discover user profiles for recommendation of POI in LBS. We first use topic modeling to learn user's interest topics from content information associated with POIs that he/she has visited. We then infer topic distribution of POI based on the learned word-topic distribution. Because the user's interest topics learnt from texts may be broadly and imprecisely, we propose a method to re-weight the user interest topics for personalized place recommendation based on extended relevance feedback. The experiments shown that the proposed recommendation algorithm achieved better performance compared to state-of-the-art algorithms on real-world LBS dataset.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have