Identifying locations from geospatial trajectories

Harnessing the latent knowledge present in geospatial trajectories allows for the potential to revolutionise our understanding of behaviour. This paper discusses one component of such analysis, namely the extraction of significant locations. Specifically, we: (i) present the Gradient-based Visit Extractor (GVE) algorithm capable of extracting periods of low mobility from geospatial data, while maintaining resilience to noise, and addressing the drawbacks of existing techniques, (ii) provide a comprehensive analysis of the properties of these visits and consequent locations, extracted through clustering, and (iii) demonstrate the applicability of GVE to the problem of visit extraction with respect to representative use-cases.