TMGCN-ATT: A Traffic Prediction Model with Multi-Hop GCN and Self-Attention Mechanisms

Accurate real-time traffic forecasting is essential for intelligent traffic systems and urban planning but remains challenging due to the complexity of road networks and dynamic, time-dependent changes. To address this, we propose TMGCN-ATT, a novel model that combines Multi-Hop Graph Convolutional Networks (MGCN) to capture spatial dependencies and Gated Recurrent Units (GRU) with a self-attention mechanism to model temporal dynamics in traffic data. The multi-hop GCN aggregates information from distant nodes, capturing higher-order spatial relationships, which the GRU then processes with self-attention to learn temporal patterns. The performance of the model is evaluated using four metrics: accuracy, root mean squared error (RMSE), mean absolute error (MAE), and the coefficient of determination (R²). Experimental results demonstrate that TMGCN-ATT consistently surpasses baseline models such as SVR, GRU, GCN, and TGCN across two urban traffic datasets, Los-Loop and SZ-Taxi. In traffic forecasting tasks, TMGCN-ATT achieves an average accuracy of 83.65% and an MAE of 2.5 across the two datasets.