Lightweight Clustering of Cell IDs into Meaningful Neighbourhoods

Abstract. Location-based services are becoming a norm in todays mo-bile phones, leveraging sensor data such as GPS, inertial sensors, and magnetic field sensors, or utilize up-to-date databases of RF beacons ’ lo-cations such as WiFi access point (AP) and cellular towers (cell ID). With the exception of cell ID-based localization, where the cellular antenna is enabled for the core functionality of the phone, the other methods require the use of extra resources, increasing the procesing and power consump-tion of the mobile phone. This paper presents a lightweight method to form semantically meaningful clusters of cell IDs from cell ID sequences, via analysis of 15 weeks of data collected from a real user.