A classification framework and computational methods for human interaction analysis using movement data

Abstract Interaction analysis for moving individuals in space and time can contribute to understanding urban dynamics and human social networks. Recent advancements in trajectory analytics have created methods to identify and extract spatiotemporal patterns of interaction using movement tracking data. However, existing definitions and classifications of interaction between moving individuals are isolated. This research is the first attempt toward developing a taxonomy of human interaction in space and time that are observable from movement data. In the proposed framework, interaction between moving individuals is divided into four categories based on the time when it occurs and the duration of the interaction, including concurrent encounter, concurrent interaction, delayed encounter, and delayed interaction. While the duration of interaction can play a significant role in classifying human interaction, techniques to quantify the duration of interaction are limited. This article incorporates an evaluation of the time‐geographic interaction analysis method to quantify the duration of interaction between moving individuals using human‐generated GPS tracking data in California. The evaluation is a comparison between the existing time‐geographic‐based approach that mainly relies on potential path area computation and polygon intersection and the proximity‐based approach (i.e., based on the intersection of spatial and temporal buffers of two individuals) that is commonly used in digital contact tracing technologies. The results show that the time‐geographic‐based approach can measure the duration of interaction more accurately, whereas the proximity‐based approach underestimates the duration of interactions.