U-Net network with fused attention mechanism for traffic flow prediction

The traffic prediction problem involves jointly modeling nonlinear spatiotemporal dependencies at different scales, but capturing the high-dimensional dynamic spatial dependencies between node locations remains a significant challenge. To address this issue, this paper proposes a U-Net network with a fused attention mechanism (FAU-Net), which utilizes the U-Net architecture to convolve the overall structure of the graph, integrating detail information from low-level channels and high-level nodes by increasing the channel depth. Additionally, a channel attention mechanism is incorporated to enhance the model's ability to focus on important features. We conducted extensive experiments on real-world traffic datasets and compared the performance of the proposed method with several baseline models, validating the effectiveness of the model.