Estimating users' home and work locations leveraging large-scale crowd-sourced smartphone data

Estimating the home and work locations of users is important for applications such as city planning and personalized recommendations. Although existing approaches can achieve a reasonable precision, they rely on fine-grained sensor data with high sampling rate. Therefore, these approaches come with a high cost and are only studied in small samples of volunteers and thus cannot benefit large-scale Internet users. In this article we propose a method to use crowd-sourced location data from mobile devices to estimate the home and work locations of large-scale users, leveraging the computation power of the cloud. Experimental results demonstrate our approach achieves a good estimation precision. Moreover, we further study how the estimated home and work locations can be used in two typical applications that are difficult problems using traditional methods but can be elegantly solved by leveraging the proposed crowd-sourced approach.