GeoMAE: Masking Representation Learning for Spatio-Temporal Graph Forecasting with Missing Values
A new research paper introduces GeoMAE, a method for spatio-temporal graph forecasting that handles missing data. Missing values in urban intelligence systems, caused by environmental conditions and equipment failures, hinder traffic and energy predictions. Existing methods neglect dynamic spatial correlations in sensor networks and struggle with complex missing patterns and varying maintenance conditions. GeoMAE addresses these issues through masking representation learning.
Key facts
- GeoMAE is a method for spatio-temporal graph forecasting with missing values.
- Missing data in urban intelligence systems is due to adverse environmental conditions and equipment failures.
- Missing data affects traffic forecasting and energy consumption prediction.
- Existing methods neglect dynamic spatial correlations in sensor networks.
- Complex missing data patterns compound the problem.
- Variability in maintenance conditions causes fluctuation in missing data ratio.
- The paper is from arXiv:2508.14083v3.
- The announcement type is replace-cross.
Entities
Institutions
- arXiv