Graph-Conditioned Mixture of Graph Neural Network Experts for Traffic Forecasting

2026-05-28

AI Summary

This work addresses the limitation of existing traffic forecasting methods that employ a single unified model while neglecting the functional, structural, and dynamic heterogeneity across different regions of a road network. To overcome this, the authors propose GC-MoE, a graph-conditioned mixture-of-experts framework that dynamically combines pretrained spatiotemporal graph neural network experts for each node, training only a lightweight routing module (approximately 17k parameters). This approach enables node-level personalized modeling by freezing expert networks and introducing a learnable routing mechanism, optionally augmented with a graph-constrained refinement layer and an ST-LoRA adapter. Evaluated on four standard benchmarks—PEMS04, PEMS07, METR-LA, and PEMS-BAY—the method achieves state-of-the-art performance, significantly reducing MAE, RMSE, and MAPE.

Abstract

Spatio-temporal forecasting on sensor graphs is commonly tackled with a single backbone architecture applied uniformly across all nodes, although graph regions can exhibit different dynamics. Road segments differ in functional class, structure, and traffic behavior, suggesting that node-wise expert specialization can be useful. We propose GC-MoE, a graph-conditioned mixture of experts framework that assigns each node a personalized combination of frozen forecasting experts based on graph topology and the recent traffic input window. GC-MoE combines frozen pretrained spatio-temporal GNN experts with an input-aware, spatially contextualized router while training only a lightweight routing module. We also study a bounded graph-conditioned output refinement layer as an optional extension and include node-adaptive ST-LoRA adapters only as an ablation diagnostic. Across four standard benchmarks (PEMS04, PEMS07, METR-LA, and PEMS-BAY), GC-MoE improves MAE over a zero-parameter ensemble baseline, with competitive RMSE and MAPE, while training only ~17K parameters on top of 1.5M frozen expert weights. The implementation is available at https://github.com/Ahghaffari/gc_moe.

Problem

• traffic forecasting
• spatio-temporal forecasting
• graph neural networks
• heterogeneous dynamics
• sensor graphs

Innovation

• Graph Neural Networks
• Mixture of Experts
• Spatio-temporal Forecasting
• Parameter-Efficient Learning
• Traffic Prediction