HyperD: Hybrid Periodicity Decoupling Framework for Traffic Forecasting

2025-11-12

AI Summary

Traffic forecasting faces two key challenges: (1) modeling complex, dynamically evolving spatial dependencies within road networks, and (2) capturing temporal patterns amid coexisting multi-scale periodicities (e.g., daily/weekly cycles) and abrupt disturbances (e.g., accidents, weather). To address these, we propose a Hybrid Periodicity Decoupling Framework—the first to semantically decouple periodic and residual components. Our approach features a dual-branch representation module coupled with a dual-view alignment loss for interpretable component separation. We further introduce learnable periodic embeddings, spatiotemporal attention, and a frequency-domain complex-valued MLP to precisely model high-frequency fluctuations and fine-grained periodic patterns in the spectral domain. Evaluated on four real-world datasets, our method achieves state-of-the-art performance, while demonstrating superior robustness to disturbances and higher computational efficiency.

Technology Category

• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Machine Learning: Time-Series/Data Streams

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding

Abstract

Accurate traffic forecasting plays a vital role in intelligent transportation systems, enabling applications such as congestion control, route planning, and urban mobility optimization.However, traffic forecasting remains challenging due to two key factors: (1) complex spatial dependencies arising from dynamic interactions between road segments and traffic sensors across the network, and (2) the coexistence of multi-scale periodic patterns (e.g., daily and weekly periodic patterns driven by human routines) with irregular fluctuations caused by unpredictable events (e.g., accidents, weather, or construction). To tackle these challenges, we propose HyperD (Hybrid Periodic Decoupling), a novel framework that decouples traffic data into periodic and residual components. The periodic component is handled by the Hybrid Periodic Representation Module, which extracts fine-grained daily and weekly patterns using learnable periodic embeddings and spatial-temporal attention. The residual component, which captures non-periodic, high-frequency fluctuations, is modeled by the Frequency-Aware Residual Representation Module, leveraging complex-valued MLP in frequency domain. To enforce semantic separation between the two components, we further introduce a Dual-View Alignment Loss, which aligns low-frequency information with the periodic branch and high-frequency information with the residual branch. Extensive experiments on four real-world traffic datasets demonstrate that HyperD achieves state-of-the-art prediction accuracy, while offering superior robustness under disturbances and improved computational efficiency compared to existing methods.

Problem

• Decoupling traffic data into periodic and residual components
• Modeling multi-scale periodic patterns with irregular fluctuations
• Handling complex spatial dependencies in traffic forecasting

Innovation

• Decouples traffic data into periodic and residual components
• Uses spatial-temporal attention for periodic patterns extraction
• Models non-periodic fluctuations with frequency-aware complex-valued MLP