Skip to Content
You are currently on the new version of our website. Access the old version .
  • Article
  • Open Access

24 July 2025

27 Pages

Multi-Channel Spatio-Temporal Data Fusion of ‘Big’ and ‘Small’ Network Data Using Transformer Networks

,
,
,
,
and
1
SpaceTimeLab, Department of Civil, Environmental and Geomatic Engineering, University College London, London WC1E 6BT, UK
2
School of Emergency Management Science and Engineering, University of Chinese Academy of Sciences, Beijing 100049, China
*
Author to whom correspondence should be addressed.
†
These authors contributed equally to this work.

Abstract

The integration of heterogeneous spatio-temporal datasets presents a critical challenge in geospatial data science, particularly when combining large-scale, passively collected “big” data with precise but sparse “small” data. In this study, we propose a novel framework—Multi-Channel Spatio-Temporal Data Fusion (MCST-DF)—that leverages transformer-based deep learning to fuse these data sources for accurate network flow estimation. Our approach introduces a Residual Spatio-Temporal Transformer Network (RSTTNet), equipped with a layered attention mechanism and multi-scale embedding architecture to capture both local and global dependencies across space and time. We evaluate the framework using real-world mobile sensing and loop detector data from the London road network, demonstrating over 89% prediction accuracy and outperforming several benchmark deep learning models. This work provides a generalisable solution for spatio-temporal fusion of diverse geospatial data sources and has direct relevance to smart mobility, urban infrastructure monitoring, and the development of spatially informed AI systems.

Article Metrics

Citations

Article Access Statistics

Multiple requests from the same IP address are counted as one view.