Graph use affects accuracy in predicting time-based data patterns
How Much Graph Is Too Much? Graph Construction and Exposure in Spatio-Temporal Forecasting
Machine Learning
Summary
Predicting things that change over time in different places, like traffic or electricity use, often uses information from nearby areas connected in a graph. This paper looks at whether and how using these connections helps prediction accuracy beyond just a location’s own history. The authors tested different ways to build these graphs and combine their information with time-based models and found some methods work better than others depending on the setting. They emphasize that both how you create and how you use the graph matter for better predictions.
What this means in practice
- •For transportation planners: Use improved graph-based forecasting to better predict traffic speed patterns across road networks for scheduling and planning.
- •For energy demand forecasters: Incorporate graphs built from residual correlations to enhance short-term electricity demand forecasts beyond single-location historical data.
Authors
Saj Maru, Qiyu Han, Krishna Mamidala
Abstract
Spatio-temporal graph neural networks combine a node's history with information from graph neighbors. The central question is whether graph neighbors provide predictive information beyond each node's own history. Forecast accuracy depends on both the choice of neighbors and the way graph messages enter the temporal model. We study these choices as graph construction and graph exposure on U.S. state unemployment, traffic speed, and electricity demand. Our diagnostic design uses two comparisons: varying exposure while holding the domain graph fixed, and varying construction while holding late gated integration fixed. A GRU using only each node's own history and randomized graph controls help assess what the graph contributes. Under a common change forecasting protocol, the same domain graph can improve or worsen error relative to the GRU across integration designs. Late integration alone is insufficient, while adding a fusion gate yields lower mean error in all three tasks. With late gated integration fixed, a graph based on residual correlations has the lowest mean error among the tested graphs, although randomized graphs perform similarly on unemployment. The findings support evaluating graph construction and integration together, with conclusions tied to the comparisons actually performed.