Extracting Meaningful User Locations from Temporally Annotated Geospatial Data

The pervasive nature of location-aware devices has enabled the collection of geospatial data for the provision of personalised services. Despite this, the extraction of meaningful user locations from temporally annotated geospatial data remains an open problem. Meaningful location extraction is typically considered to be a 2-step process, consisting of visit extraction and clustering. This paper evaluates techniques for meaningful location extraction, with an emphasis on visit extraction. In particular, we propose an algorithm for the extraction of visits that does not impose a minimum bound on visit duration and makes no assumption of evenly spaced observation.