Efficient Traffic Prediction at Scale: A Systematic Study of STGCN Architectural Depth
2026-06-08 • Machine Learning
Machine Learning
AI summaryⓘ
The authors studied a popular traffic prediction model called STGCN to see if it uses more parts (blocks) than needed. They tested versions with one, two, and three blocks on different traffic datasets. They found that the one-block model works best for short-term forecasts and is much faster and cheaper to run than the usual two-block version, which does not improve results enough to justify its extra computing cost. The three-block model is even more expensive but gives very little accuracy gain. This suggests that simpler models might be better for real-world use and research comparisons.
Spatio-temporal graph neural networksSTGCNtraffic predictionmodel complexityefficiencyinference latencygraph convolutionITS deploymentshort-term prediction
Authors
Soban Nasir Lone, Mohamed Abouelela, Taeyoung Yu, Jiwon Kim, Constantinos Antoniou
Abstract
Spatio-temporal graph neural networks (STGNNs) have become the dominant approach for traffic prediction, yet their computational requirements pose challenges for practical deployment in intelligent transportation systems (ITS). While recent work has proposed efficient alternatives to STGNNs, a fundamental question remains unexplored: are these architectures themselves over-parameterised? We examine this question using the Spatio-Temporal Graph Convolutional Network (STGCN), one of the most widely adopted models in this domain. Through systematic experiments across four diverse traffic datasets, we compare 1-block, 2-block (standard), and 3-block STGCN variants. Our findings reveal that the single-block architecture achieves optimal performance for short-term prediction (10 mins) on three of four datasets, while incurring only marginal degradation ($\leq$1.8% relative error) at longer horizons. Crucially, the 2-block variant incurs 61% higher CPU inference latency and 37% lower throughput relative to 1-block -- substantial overhead for resource-constrained ITS deployment. The 3-block architecture offers no favourable tradeoff, more than doubling computational cost for $<$0.5% relative improvement. These results suggest that the default 2-block STGCN may be over-parameterised for many applications, with implications for both practitioners deploying traffic prediction systems and researchers benchmarking efficiency-focused methods.