Exploring multilevel regularity in human mobility patterns using a feature engineering approach: A case study in Chicago

Fine-grained trajectory data offers the opportunity to advance our understanding of regularity in individual mobility choices. Existing research on regular location-based mobility patterns cannot fully capture the complexity of day-to-day individual trajectories that are critical for downstream predictive models. To bridge the gap, we construct a comprehensive mobility profile using interpretable features to quantify the regularity of human mobility from three levels, e.g. locations, daily itineraries, and routes. An empirical study uses over 93k trips from 776 car drivers in the Chicago metropolitan area validates the routinely regular patterns in users' mobility choices. A feature engineering approach is then designed for user segmentation. Six user clusters are discovered from the users' mobility profiles. The clusters exhibit heterogeneous commuting behavior and preference for motif choices. The improved multi-level understanding of repeated travel behavior can further assist transportation modeling and planning.