Modeling checkpoint-based movement with the earth mover’s distance

Movement data comes in various forms, including trajectory data and checkpoint data. While trajectories give detailed information about the movement of individual entities, checkpoint data in its simplest form does not give identities, just counts at checkpoints. However, checkpoint data is of increasing interest since it is readily available due to privacy reasons and as a by-product of other data collection. In this paper we propose to use the Earth Mover’s Distance as a versatile tool to reconstruct individual movements or flow based on checkpoint counts at different times. We analyze the modeling possibilities and provide experiments that validate model predictions, based on coarse-grained aggregations of data about actual movements of couriers in London, UK. While we cannot expect to reconstruct precise individual movements from highly granular checkpoint data, the evaluation does show that the approach can generate meaningful estimates of object movements.