16 April 2025 in IEEE Internet of Things Journal
Accurate traffic prediction is crucial for intelligent transportation systems (ITS). However, recent advancements in deep learning have encountered significant challenges in effectively capturing long-range dependencies and modeling complex intervariable relationships, particularly under real-time processing constraints. These limitations primarily arise from the inherent tradeoff in current model architectures: while some are optimized for short-term forecasting with constrained receptive fields, others prioritize long-term prediction accuracy at the cost of computational efficiency. Moreover, the dynamic and intricate interactions among nodes present significant scalability challenges, increasingly hindering efficient modeling as the road network grows. To address these limitations, we propose temporal window and sparse spatial attention transformer (TWIST), a novel architecture specifically designed for traffic forecasting. The proposed framework incorporates two key innovations: 1) a trend-aware window attention mechanism in the temporal dimension that effectively and efficiently captures multiscale temporal dynamics and 2) a sparse attention mechanism in the spatial dimension that optimizes computational efficiency by selectively focusing on significant spatial nodes. Extensive experiments on eight real-world datasets demonstrate that TWIST consistently outperforms state-of-the-art methods across various scenarios, including regular, long-term, and large-scale traffic forecasting tasks, while maintaining competitive computational efficiency. Our code is publicly available at https://github.com/STGTraffic/TWIST.
26 November 2025 in IEEE Internet of Things Journal
Accurate traffic flow forecasting is crucial in intelligent transportation systems (ITSs) for mitigating congestion and enhancing intelligent traffic management. Although recent methods have shown excellent performance, they still encounter two major limitations: 1) existing models face challenges in balancing prediction accuracy and computational efficiency and 2) existing spatial modeling methods use unified mechanisms to model all nodes, neglecting the inherent heterogeneity among them. To address these limitations, we propose a spatiotemporal architecture, TA-CSA for traffic prediction, which consists of temporal aggregation pooling (TAP) and clustered spatial attention (CSA). Specifically, in the temporal dimension, we design the TAP module. This module extracts temporal features from multiple time windows through parallel pooling operations, thereby strengthening the model’s capacity to capture global temporal trends and dynamic fluctuations. In the spatial dimension, we introduce CSA. The method employs hierarchical clustering to divide nodes with similar functional characteristics into the same cluster and performs local attention calculations to achieve regional-level spatial perception. Extensive experiments on six real-world traffic datasets demonstrate that TA-CSA consistently outperforms state-of-the-art baselines in both routine and long-horizon forecasting tasks.
1 January 2026 in IEEE Internet of Things Journal