Short-term traffic flow prediction of coastal urban road nodes based on A3-TGCN

In recent years, with the improvement in living standards and economic development, the number of cars in the city has been increasing year by year. This has put pressure on the carrying capacity of the roads and led to traffic congestion. However, it is difficult for traditional time series models to predict future traffic flows accurately. In this paper, we take the traffic data of the coastal urban area as our research object, combining spatial and temporal factors of the city. We observed the excellent performance of the improved model in complex environments by the attention temporal graph convolutional network(A3T-GCN) model. The root mean square error(RMSE) value is 33.305, the mean absolute error (MAE) value is 21.257, and the variance(var) is 90.3%. This study provides a practical algorithmic foundation for intelligent transport. The A3T-GCN algorithm, which is based on the self-attention mechanism, has a significant advantage in traffic flow prediction, especially when there is limited data