Related Experiment Video
Updated: Jun 14, 2026
05:55
Modeling the Functional Network for Spatial Navigation in the Human Brain
Published on: October 13, 2023
Transferable human mobility network reconstruction with neuroGravity
Jinming Yang1,2, Shaoyu Huang1, Zongyuan Huang1
1MoE Key Laboratory of Artificial Intelligence, Institute of AI, School of Computer Science, Shanghai Jiao Tong University, Shanghai, China.
Nature Computational Science
|June 12, 2026
Summary
NeuroGravity, a deep learning model, reconstructs human mobility networks from limited data, offering insights into urban planning and public health. Spatial income segregation impacts model accuracy, enabling better data generation for underserved regions.
Area of Science:
- Urban Science
- Computational Social Science
- Deep Learning
Background:
- Accurate human mobility modeling is vital for urban planning and public health.
- Undeveloped regions often lack comprehensive travel survey data, hindering mobility network analysis.
- Reconstructing mobility from publicly available data is crucial for these areas.
Purpose of the Study:
- To develop a deep learning model (neuroGravity) for reconstructing human mobility flows from limited observations.
- To assess the model's ability to transfer knowledge to unobserved cities.
- To identify factors influencing model transferability and generate mobility proxies for data-scarce regions.
Main Methods:
- Developed neuroGravity, a physics-informed deep learning model.
- Utilized urban facility and population distributions as input data.
- Analyzed the impact of spatial income segregation on model transferability and developed a segregation index.
Main Results:
- NeuroGravity reliably reconstructs mobility flows from limited data.
- Regional representations from neuroGravity correlate with socioeconomic and livability status, serving as survey proxies.
- Spatial income segregation significantly influences model transferability, which can be predicted by a novel index.
- Generated mobility flow proxies for over 1,200 cities worldwide.
Conclusions:
- NeuroGravity offers a scalable solution for reconstructing human mobility in data-scarce regions.
- Spatial income segregation is a key factor for successful model transferability in mobility network reconstruction.
- The model has significant potential to address data shortages in resource-limited and underdeveloped areas for urban planning and public health initiatives.
