A data synthesis approach for human daily activities and trajectories.

What is it about?

In this paper, we propose to leverage the power of deep learning for synthesizing human mobility data. Specifically, we design a generative adversarial-based framework, ActSTD, effectively incorporating the spatiotemporal dynamics with continuous flow and instantaneous updates in time and space. Extensive experiments demonstrate the superiority of the proposed framework in generating high-quality activity trajectories.

Featured Image

Photo by NASA on Unsplash

Photo by NASA on Unsplash

Why is it important?

Generating activity trajectories is of great importance for many applications, especially for epidemic modeling. Despite the great value of individual activity trajectories in pan-demic control, such data are highly limited in applications due to privacy issues and commercial concerns. Instead, realistic simulation of individual-level daily activities to generate massive high-quality activity trajectories becomes an essential and meaningful research problem, which covers the deficiency of real-world data in modeling the pandemic spread and facilitating rational policymaking.