Discovering significant places from mobile phones - A mass market solution

Abstract

In this paper we propose a mass market solution on mobile phones to discover a user's significant places solely from observed cell IDs. It does not require either cell-ID-to-physical-location mapping or the capability of obtaining multiple cell IDs on the phone simultaneously, and is able to run on virtually any mobile phone today. Our solution is centered around a cell ID clustering algorithm based on temporal correlations. It is able to prevent over-clustering and handles missing data well. We evaluate the solution with real-life data that the author has collected over a period of eight weeks. Results show that we are able to discover not only places of utter importance, but also certain less frequently recurring places and one-time travel destinations that bear significance in one's life. © 2009 Springer Berlin Heidelberg.