Adaptive Graph Pruning with Sudden-Events Evaluation for Traffic Prediction using Online Semi-Decentralized ST-GNNs

2025-12-19

AI Summary

To address the high communication overhead in edge-based intelligent traffic forecasting caused by redundant transmission of neighborhood features across cloud nodes in spatio-temporal graph neural networks (ST-GNNs), this paper proposes an online semi-decentralized ST-GNN framework. Our method introduces: (1) an adaptive graph pruning mechanism that dynamically adjusts the pruning rate based on real-time model performance, balancing spatial information preservation and communication load; (2) SEPA—a novel, event-aware evaluation metric specifically designed to quantify model responsiveness to sudden congestion onset and recovery; and (3) integration of federated learning paradigms, including serverless and gossip-based approaches, to enhance collaborative efficiency. Experiments on PeMS-BAY and PeMSD7-M demonstrate significant reductions in communication cost across all prediction horizons, with no degradation in forecasting accuracy. SEPA analysis confirms the critical role of dynamic graph topology in modeling abrupt traffic events.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Search and Optimization: Distributed Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Systems and Infrastructure for Web, Mobile and WoT: Federated Web and WoT systems, including distributed, federated and edge-based data processing

Abstract

Spatio-Temporal Graph Neural Networks (ST-GNNs) are well-suited for processing high-frequency data streams from geographically distributed sensors in smart mobility systems. However, their deployment at the edge across distributed compute nodes (cloudlets) createssubstantial communication overhead due to repeated transmission of overlapping node features between neighbouring cloudlets. To address this, we propose an adaptive pruning algorithm that dynamically filters redundant neighbour features while preserving the most informative spatial context for prediction. The algorithm adjusts pruning rates based on recent model performance, allowing each cloudlet to focus on regions experiencing traffic changes without compromising accuracy. Additionally, we introduce the Sudden Event Prediction Accuracy (SEPA), a novel event-focused metric designed to measure responsiveness to traffic slowdowns and recoveries, which are often missed by standard error metrics. We evaluate our approach in an online semi-decentralized setting with traditional FL, server-free FL, and Gossip Learning on two large-scale traffic datasets, PeMS-BAY and PeMSD7-M, across short-, mid-, and long-term prediction horizons. Experiments show that, in contrast to standard metrics, SEPA exposes the true value of spatial connectivity in predicting dynamic and irregular traffic. Our adaptive pruning algorithm maintains prediction accuracy while significantly lowering communication cost in all online semi-decentralized settings, demonstrating that communication can be reduced without compromising responsiveness to critical traffic events.

Problem

• Reduces communication overhead in distributed ST-GNNs for traffic prediction
• Adaptively prunes redundant neighbor features to preserve spatial context
• Introduces a metric to assess responsiveness to sudden traffic events

Innovation

• Adaptive pruning algorithm dynamically filters redundant neighbor features
• Adjusts pruning rates based on recent model performance for accuracy
• Introduces SEPA metric to measure responsiveness to traffic events