AN INTEGRATED APPROACH TO LOCATION MODELING

When smartphone users are excited with the powerful applications on their iPhones and Androids, they might have little idea that these advanced cell phones enable easy tracking of their locations recorded with time, which could lead to serious privacy and security problems. While research has found that even anonymized location records are prone to cracking, location modeling provides a platform for privacy protection through processing the position information.

A location is a place on the Earth's surface treated as 2-dimensional, flat, and large-scale space called geographic space, which this research considers as having two aspects: the geometric space and the name space. Each location is an (geographic) entity that has a footprint in the geometric space and a placename in the name space. A footprint is treated as a polygon determined by geographic coordinates, while a placename is simply a word or phrase. There are inclusion relations between individual footprints as well as their aggregations. Footprints aggregated by inclusion relations form structures made up of levels in the geometric space, and so do the corresponding placenames in the name space. Corresponding aggregations of footprints and placenames bring about aggregations of entities, which are stored with gazetteer models presented in this thesis.

Location information contains two types of coarseness, which are carried by granularity (e.g., 50–100km2 in area) for footprints and (geographic) class (e.g., county) for placenames, respectively. One purpose of location modeling is to provide consistent switching to a designated level of coarseness, which is enabled by using each level in a structure as a granularity or class, specifically, in the form of level–granularity consistency for footprints and level–class consistency for placenames.

The original structures of a gazetteer, however, only provide level–class consistency. This research thus develops an augmentation mechanism to enlarge an original gazetteer with virtual entities, resulting in an augmented gazetteer, which guarantees both level–class consistency and level–granularity consistency through a pair of augmented structures. The approach then provides an integrated framework consisting of such a pair of structures, which enables conversions between the footprints and placenames through their associations in the entities.