Table 3 Performance comparison of different approaches on four datasets.
From: Multi-scale spatio-temporal graph neural network for urban traffic flow prediction
PEMS03 | PEMS04 | PEMS07 | PEMS08 | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
RMSE | MAE | MAPE | RMSE | MAE | MAPE | RMSE | MAE | MAPE | RMSE | MAE | MAPE | |
STG-NCDE | 27.09 | 15.57 | 15.06 | 31.09 | 19.21 | 12.76 | 33.84 | 20.53 | 8.80 | 24.81 | 15.45 | 9.92 |
ST-aware | 26.63 | 15.17 | 15.83 | 31.02 | 19.06 | 12.52 | 34.05 | 20.74 | 8.77 | 24.62 | 15.41 | 9.94 |
DSTAGNN | 27.21 | 15.57 | 14.68 | 31.46 | 19.30 | 12.70 | 34.51 | 21.42 | 9.01 | 24.77 | 15.67 | 9.94 |
Auto-DSTSGN | 25.17 | 14.59 | 14.22 | 30.48 | 18.85 | 13.21 | 33.02 | 20.08 | 8.57 | 23.76 | 14.74 | 9.45 |
PDFormer | 25.39 | 14.94 | 15.82 | 29.97 | 18.32 | 12.10 | 32.87 | 19.83 | 8.53 | 23.51 | 13.58 | 9.05 |
STGSA | 27.89 | 15.36 | 14.45 | 31.30 | 19.32 | 12.90 | 34.30 | 20.80 | 8.86 | 24.28 | 15.26 | 9.81 |
ISTNet | 25.14 | 15.12 | 15.43 | 30.46 | 18.54 | 12.52 | 33.06 | 19.79 | 8.77 | 23.39 | 14.13 | 9.43 |
DeepSTUQ | 26.77 | 15.13 | 14.03 | 31.68 | 19.11 | 12.71 | 33.71 | 20.36 | 8.63 | 24.60 | 15.44 | 10.06 |
STJGCN | 25.70 | 14.92 | 14.81 | 30.35 | 18.81 | 11.92 | 33.01 | 19.95 | 8.31 | 23.74 | 14.53 | 9.15 |
DEC-Former | 23.55 | 14.33 | 14.27 | 29.24 | 18.23 | 12.04 | 33.04 | 19.48 | 8.54 | 23.06 | 13.23 | 9.12 |
STD-MAE | 24.43 | 13.80 | 13.96 | 29.25 | 17.80 | 11.97 | 31.44 | 18.65 | 7.84 | 22.47 | 13.44 | 8.76 |
STGMS | 22.22 | 12.87 | 13.17 | 24.48 | 14.89 | 10.86 | 24.09 | 14.61 | 6.78 | 17.52 | 10.87 | 7.61 |
Improved(%) | 9.05 | 6.75 | 5.66 | 16.28 | 16.35 | 8.89 | 23.38 | 21.66 | 13.52 | 22.03 | 17.84 | 13.13 |
