<p>Traffic prediction plays a crucial role in intelligent transportation systems. However, the complexity of traffic data, stemming from a mix of local and global spatial–temporal correlations, poses significant challenges for accurate traffic flow prediction. Most recurrent neural networks or convolutional neural networks struggle to effectively capture temporal dependencies, especially long-term ones. Additionally, existing graph neural networks typically use static graph structures to capture spatial dependencies, often overlooking the dynamic patterns inherent in traffic networks. In this paper, we propose a novel spatial–temporal adaptive dynamic graph convolutional network for traffic flow prediction. This network incorporates a Gated Temporal Convolution Network (Gated TCN) that uses dilated causal convolutions at different granular levels to capture temporal dependencies in traffic flow. Additionally, we designed a spatial static–dynamic graph learning layer, which integrates static adaptive graph learning, dynamic graph learning, and a spatial gated fusion module to synchronously capture dynamic spatial–temporal features from historical traffic data. Experiments conducted on two real-world traffic datasets demonstrate that the proposed model effectively extracts dynamic spatial–temporal features and significantly outperforms popular baseline methods in prediction accuracy.</p>