ADMFormer: An Adaptive-Decomposition Transformer with Time-Varying Masked Spatial Attention for Traffic Forecasting
2026-05-25 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors focus on predicting traffic by addressing two main challenges: traffic patterns have regular trends but also unpredictable events, and the connections between locations change over time and are not always simple. They designed a model called ADMFormer that splits traffic data into regular and fluctuating parts, then studies these separately to better understand their patterns. Their model also smartly chooses which location connections to pay attention to based on current traffic, avoiding distractions from less useful data. Tests on real traffic data show their approach works better than previous methods.
traffic forecastingtemporal patternsspatial dependenciestransformer modelsadaptive gatingtime-varying attentionperiodic regularitiesevent-driven fluctuationsmasked spatial attentiondual-branch temporal module
Authors
Ruiwen Gu, Qitai Tan, Yahao Liu, Xiao-Ping Zhang
Abstract
Accurate traffic forecasting is essential for intelligent transportation systems, supporting a wide range of real-world applications. However, it remains challenging due to two key factors:~(1) Traffic series contain heterogeneous temporal patterns, where stable periodic regularities coexist with event-driven fluctuations. Existing methods often treat them within a unified representation, limiting their ability to capture fine-grained temporal dynamics.~(2)Spatial dependencies among nodes are inherently dynamic and sparse, while dense all-pairs attention often introduces redundant interactions and amplifies noise. To address these issues, we propose ADMFormer, an Adaptive-Decomposition Transformer with Time-Varying Masked Spatial Attention. Specifically, ADMFormer first employs a time-node adaptive gating mechanism to decouple traffic signals into dominant regularities and residual fluctuations that vary across time and nodes. A dual-branch temporal module is then designed to separately capture global periodic dependencies and high-frequency irregular variations from these two decomposed components. Furthermore, ADMFormer introduces a time-varying masked spatial attention that sparsifies spatial interactions based on real-time traffic states, thereby effectively preserving dynamic and informative dependencies. Extensive experiments on four real-world datasets demonstrate that ADMFormer achieves state-of-the-art performance.