Table 2 Comparison of traffic flow prediction performance between STCMFA and baseline models.
From: Spatial–temporal combination and multi-head flow-attention network for traffic flow prediction
Model | PeMS03 | PeMS04 | PeMS07 | PeMS08 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
MAE | MAPE (%) | RMSE | MAE | MAPE (%) | RMSE | MAE | MAPE (%) | RMSE | MAE | MAPE (%) | RMSE | |
LSTM | 21.33 | 23.33 | 35.11 | 27.14 | 18.20 | 41.59 | 29.98 | 13.20 | 45.84 | 22.20 | 14.20 | 34.06 |
DCRNN | 18.18 | 18.91 | 30.31 | 24.70 | 17.12 | 38.12 | 25.30 | 11.66 | 38.58 | 17.86 | 11.45 | 27.83 |
STGCN | 17.49 | 17.15 | 30.12 | 22.70 | 14.59 | 35.55 | 25.38 | 11.08 | 38.78 | 18.02 | 11.40 | 27.83 |
ASTGCN(r) | 17.69 | 19.40 | 29.66 | 22.93 | 16.56 | 35.22 | 28.05 | 13.92 | 42.57 | 18.61 | 13.08 | 28.16 |
STG2Seq | 19.03 | 21.55 | 29.73 | 25.20 | 18.77 | 38.48 | 32.77 | 20.16 | 47.16 | 20.17 | 17.32 | 30.71 |
Graph WaveNet | 19.85 | 19.31 | 32.94 | 25.45 | 17.29 | 39.70 | 26.85 | 12.12 | 42.78 | 19.13 | 12.68 | 31.05 |
STSGCN | 17.48 | 16.78 | 29.21 | 21.19 | 13.90 | 33.65 | 24.26 | 10.21 | 39.03 | 17.13 | 10.96 | 26.80 |
STGODE | 16.50 | 16.69 | 27.84 | 20.84 | 13.77 | 32.82 | 22.99 | 10.14 | 37.54 | 16.81 | 10.62 | 25.97 |
STDSGNN | 16.12 | 16.15 | 25.59 | 20.67 | 13.83 | 32.40 | 22.91 | 10.06 | 34.95 | 16.73 | 10.84 | 25.59 |
STGPCN (Kronecker) | 17.11 | 16.48 | 28.99 | 20.96 | 13.78 | 33.35 | 24.02 | 10.08 | 38.77 | 16.41 | 10.43 | 25.60 |
LEISN-ED | 15.83 | 14.66 | 26.05 | – | – | – | – | – | – | 15.94 | 10.18 | 24.96 |
STSGRU | 15.45 | 15.85 | 24.13 | 20.11 | 13.86 | 31.80 | 21.50 | 9.08 | 34.40 | 15.68 | 10.67 | 25.12 |
STCMFA(our) | 15.48 | 13.52 | 22.91 | 19.78 | 12.51 | 29.51 | 21.34 | 8.39 | 33.39 | 16.09 | 9.27 | 24.12 |
