For Here or to Go: Predicting Client Dwell Time in WiFi Hotspots
Modern mobile phones provide a plethora of sensors, enabling high-resolution measurement of the user's behavior. Further, user behaviors are not random-detailed measurements of current activities can be suggestive of the near future. The authors envision that such behavioral insights can help deliver enhanced network services. While behavior-aware networking is a broad research theme, one possibility is to predict how long someone will remain at a WiFi hotspot, her dwell time. In this paper, they design a general framework for capturing relevant sensor data and use it for live predictions of dwell time for associated clients. They believe the implications of such a system are numerous.