Fig 1.
CNN structure.
Fig 2.
CBAM structure.
Fig 3.
MSCALSTM.
Fig 4.
LSTM.
Table 1.
Performance comparison on traffic datasets of different sizes.
Table 2.
Performance comparison on traffic datasets of different scales.
Table 3.
Average performance comparison of different methods on PeMSD3 and PeMSD4.
Table 4.
Average performance comparison of different methods on PeMSD7 and PeMSD8.
Fig 5.
Comparison on Pemsd3.
Fig 6.
Comparison on Pemsd4.
Fig 7.
Comparison on Pemsd7.
Fig 8.
Comparison on Pemsd8.
Table 5.
Comparison of AIC and BIC of different methods on four datasets.
Table 6.
Comparison of average performance of different methods on the energy dataset.
Table 7.
Comparison of AIC and BIC of different methods on energy dataset.
Table 8.
Ablation experiments at PeMSD3 and PeMSD4.
Table 9.
Ablation experiments at PeMSD7 and PeMSD8.
Table 10.
Ablation experiments on four datasets.
Fig 9.
Ablation on Pemsd3.
Fig 10.
Ablation on Pemsd4.
Fig 11.
Ablation on Pemsd7.
Fig 12.
Ablation on Pemsd8.
Table 11.
Ablation experiments on Energy dataset.
Fig 13.
Real and predicted values on Pemsd4.
Fig 14.
Local amplification of true and predicted values on Pemsd4.
Fig 15.
Real and predicted values on Pemsd8.
Fig 16.
Local amplification of true and predicted values on Pemsd8.
Fig 17.
Attention weight heat map.
Table 12.
MAE significance comparison between MSCALSTM and baseline models based on PeMSD3 dataset.
Table 13.
MAE significance comparison between MSCALSTM and baseline models based on PeMSD4 dataset.
Table 14.
MAE significance comparison between MSCALSTM and baseline models based on PeMSD7 dataset.
Table 15.
MAE significance comparison between MSCALSTM and baseline models based on PeMSD8 dataset.
Table 16.
MAE significance comparison between MSCALSTM and baseline models based on Energy dataset.
Fig 18.
Computational process.