On Physical-Social-Aware Localness Inference by Exploring Big Data from Location-Based Services

A user's localness (i.e., whether a user is a local resident in a city or not) and a venue's local attractiveness (i.e., the likelihood of a venue to attract local people) are important information for many location-based applications related with Cyber-Physical Systems (CPS), such as participatory sensing, urban planning, traffic control and localized travel recommendations. Previous effort has been devoted to geo-locating users in a city using supervised learning approaches, which depend on the availability of high quality training datasets. However, it is difficult to obtain such training datasets in the real-world CPS applications due to the issue of privacy. In this work, we develop an unsupervised approach, called a Physical-Social-Aware Inference (PSAI) scheme, to jointly infer a user's localness and a venue's local attractiveness by exploring both the physical and social information embedded in the location-based social networks (LBSN). We further implement a parallel PSAI framework on the platform of a Graphic Processing Unit (GPU) to enhance its ability to process large-scale data. Our extensive experiments on the real-world LBSN datasets demonstrate the effectiveness and efficiency of the PSAI scheme compared to the state-of-the-art baselines.