TSFR: Time-aware Semantic-based Friend Recommendation in LBSNs

Abstract

Advances in broadband wireless networks and location sensing technologies have led to the emergence of location-based online social networks (LBSNs) in recent years. Users' passion for sharing locations has attracted much attention to traditional social networks. Therefore, the great amount of check-in data can be used to make recommendations for interesting places and to make friends. Because both semantic and time information on check-in data reflect preference and interests of users, we take both of them into consideration and propose a time-aware semantic-based recommender system in this paper. We use the Term Frequency-Inverse Document Frequency (TF-IDF) model and the Kullback-Leibler divergence (K-L divergence) to combine the semantic and time information of check-in data to make friend recommendations. To evaluate our recommender system, we get a dataset of Gowalla and build a system using the Collaborative Filtering recommender system structure. The experiment results show that our system, with the consideration of time and semantic information of check-in data, outperforms the classic collaborative filtering recommender system.