On estimating the predictability of human mobility: the role of routine

Abstract Given the difficulties in predicting human behavior, one may wish to establish bounds on our ability accurately perform such predictions. In case of mobility-related there exists a fundamental technique estimate predictability an individual’s mobility, as expressed given dataset. Although useful several scenarios, this focused mobility monolithic entity, which poses challenges understanding different types behavior that be hard predict. paper, we propose study terms two components mobility: routine and novelty, where is related preferential returns, novelty exploration. Viewing one’s these allows us identify important patterns about mobility. Additionally, argue component predict if rely history visited locations (as does), therefore here focus analyzing what affects routine. To end, (i) quantify effect predictability, (ii) gauge how much deviates from reference completely predictable, estimating amount hard-to-predict Finally, previously proposed metrics, well newly one, understand person’s Our experiments show metrics are able capture most variability (adjusted $R^{2}$ R 2 up 84.9% 96.0% GPS CDR datasets, respectively), can largely explained by three patterns: stationary patterns, person stays her current location for time period, regular visits, people visit few preferred with occasional visits other places, (iii) diversity trajectories, change order they certain locations.