Table 4 Comparison of different methods for rush hours on the HZMetro dataset.
From: TSTA-GCN: trend spatio-temporal traffic flow prediction using adaptive graph convolution network
Time | 15 min | 30 min | 60 min | ||||||
|---|---|---|---|---|---|---|---|---|---|
Metric | RMSE | MAE | MAPE | RMSE | MAE | MAPE | RMSE | MAE | MAPE |
HA | 65.53 | 40.63 | 11.51% | 67.89 | 42.08 | 11.58% | 67.22 | 40.72 | 13.21% |
RF | 84.33 | 52.07 | 15.24% | 108.25 | 65.97 | 17.56% | 136.08 | 75.40 | 20.81% |
MLP | 57.39 | 35.77 | 10.96% | 62.25 | 37.58 | 10.80% | 61.81 | 36.13 | 12.16% |
LSTM | 57.10 | 35.77 | 9.99% | 59.03 | 36.45 | 10.07% | 57.35 | 34.19 | 11.23% |
GRU | 56.31 | 35.23 | 10.12% | 58.81 | 36.59 | 10.10% | 57.14 | 34.01 | 11.08% |
ASTGCN | 60.72 | 36.82 | 11.77% | 58.30 | 35.48 | 12.15% | 59.23 | 33.59 | 13.68% |
STG2Seq | 53.28 | 35.03 | 10.73% | 56.26 | 36.96 | 10.95% | 57.69 | 35.64 | 12.25% |
DCRNN | 54.17 | 35.08 | 10.37% | 58.27 | 37.48 | 10.69% | 59.52 | 36.27 | 11.94% |
GCRNN | 55.51 | 35.68 | 10.36% | 57.34 | 37.31 | 10.54% | 58.88 | 35.94 | 11.93% |
GWN | 56.98 | 37.19 | 10.84% | 59.71 | 38.94 | 11.04% | 59.96 | 37.49 | 12.35% |
PVCGN | 49.79 | 32.63 | 9.72% | 51.63 | 33.30 | 9.52% | 51.09 | 31.43 | 10.43% |
MGT | 49.36 | 31.98 | 9.23% | 53.01 | 33.56 | 9.32% | 53.21 | 32.49 | 10.66% |
TSTA-GCN | 47.79 | 31.12 | 8.97% | 49.81 | 32.19 | 9.08% | 50.63 | 31.28 | 10.27% |
