Accurate traffic flow prediction is pivotal for intelligent transportation systems; yet, existing spatial-temporal graph neural networks (STGNNs) struggle to jointly capture the long-term structural stability, short-term dynamics, and multi-scale temporal patterns of road networks. To address these shortcomings, we propose FISTGCN, a Frequency-Aware Interactive Spatial-Temporal Graph Convolutional Network. FISTGCN enriches raw traffic flow features with learnable spatial and temporal embeddings, thereby providing comprehensive spatial-temporal representations for subsequent modeling. Specifically, it utilizes an interactive dynamic graph convolutional block that generates a time-evolving fused adjacency matrix by combining adaptive and dynamic adjacency matrices. It then applies dual sparse graph convolutions with cross-scale interactions to capture multi-scale spatial dependencies. The gated spectral block projects the input features into the frequency domain and adaptively separates low- and high-frequency components using a learnable threshold. It then employs learnable filters to extract features from different frequency bands and adopts a gating mechanism to adaptively fuse low- and high-frequency information, thereby dynamically highlighting short-term fluctuations or long-term trends. Extensive experiments on four benchmark datasets demonstrate that FISTGCN delivers state-of-the-art predictive accuracy while maintaining competitive computational efficiency.
Traffic flow prediction is a vital component of Intelligent Transportation Systems (ITSs), playing a key role in proactive traffic management and the optimization of urban mobility. However, the complex spatial–temporal dependencies, dynamic variations, and external factors in traffic networks present significant challenges for accurate predictions. In this paper, we propose MMHFormer, a novel multi-source, multi-view hierarchical Transformer model specifically designed for traffic flow prediction. MMHFormer incorporates three key innovations: (1) a multi-source gated embedding layer that integrates diverse multidimensional inputs, including spatial Laplacian embeddings, temporal periodic embeddings, and traffic occupancy embeddings, to better capture the complex dynamics of traffic conditions; (2) a hierarchical multi-view spatial attention module that models global, local, and dynamic similarity-based spatial dependencies, effectively addressing the spatial heterogeneity of traffic flows; (3) a hierarchical two-stage temporal attention mechanism that captures global temporal dependencies while adapting to node-specific temporal variations. Extensive experiments conducted on four benchmark traffic datasets demonstrate that MMHFormer outperforms state-of-the-art methods, achieving significant improvements in prediction accuracy.
Traffic flow prediction plays a crucial role in improving traffic management efficiency and optimizing traffic resource allocation. However, existing traffic flow prediction models are often difficult to accurately capture complex spatial–temporal dependencies. To this end, this paper proposes a spatial–temporal fusion Transformer based graph convolutional network (STFFormer-GCN) for traffic flow prediction, aiming to address the issues of long- and short-term dynamic spatial–temporal relationships, traffic propagation delays, and spatial–temporal feature fusion. First, we propose a multi-scale time convolution module employing recursion and sliding windows to model long- and short-term temporal dependencies. Second, a new dynamic relational convolution matrix is proposed for learning dynamic spatial dependencies that change over time and road structure. Then, we propose a time-delay cycle module that combines the current traffic flow and spatially propagated time delay information to model the propagation process of traffic flow in space. Finally, we employ Transformer's encoder to fuse dynamic spatial–temporal features with time delayed spatial–temporal features. We conducted extensive experiments on four real datasets and the results show that STFFormer-GCN achieves state-of-the-art performance,especially the MAPE value of the PeMS08 dataset experiment is improved by 4.05%. In addition, we conducted an ablation study and a single-step prediction performance study to evaluate the contribution of each component to the prediction accuracy of the model.