Fig 1.
Pearson heat map of the network traffic flow data of the Milan city network.
Fig 2.
Relationship between models.
Fig 3.
Schematic diagram of a CNN.
Fig 4.
Schematic diagram of the causal convolution structure.
Fig 5.
Schematic diagram of the expansion convolution structure.
Fig 6.
Schematic diagram of the TCN residual module.
Fig 7.
Schematic diagram of the CBAM block structure.
Fig 8.
Schematic diagram of the CAM module structure.
Fig 9.
Schematic diagram of the SAM module structure.
Fig 10.
Schematic diagram of the Transformer structure.
Fig 11.
Schematic diagram of the Transformer self-attention mechanism.
Fig 12.
Overall structure of the CSTCN-Transformer.
Fig 13.
Area division map of Milan.
Table 1.
Spatio-temporal features of dataset.
Table 2.
Hyperparameters of CSTCN-Transformer (STCN part).
Table 3.
Hyperparameters of CSTCN-Transformer (CBAM pooling part).
Table 4.
Hyperparameters of CSTCN-Transformer (CBAM convolution part).
Table 5.
Hyperparameters of CSTCN-Transformer (MLP part).
Table 6.
Hyperparameters of CSTCN-Transformer (Transformer part).
Fig 14.
Prediction effect of CSTCN-Transformer.
Table 7.
Performance comparison with baseline models.
Fig 15.
Comparison of results from baseline models.
Fig 16.
Comparison of results from ablation experiment models.
Table 8.
Performance comparison with ablation experiment models.
Table 9.
Comparison of real-time prediction and long-term prediction performance.