Table 5 Test performance with normalized RMSE and de-normalized MAE (mph).
From: 6G conditioned spatiotemporal graph neural networks for real time traffic flow prediction
Model | RMSE (norm.) | MAE (mph) | Latency (ms) |
|---|---|---|---|
ST-GCN | \(0.0719 \pm 0.0008\) (0.0703–0.0734) | \(2.131 \pm 0.041\) (2.05–2.20) | 62.7 |
ST-GAT | \(0.0713 \pm 0.0010\) (0.0695–0.0732) | \(2.081 \pm 0.038\) (2.01–2.15) | 113.3 |
Graph WaveNet | \(0.0468 \pm 0.0012\) (0.0445–0.0491) | \(1.214 \pm 0.029\) (1.16–1.27) | 49.5 |
AGCRN | \(0.0419 \pm 0.0014\) (0.0393–0.0445) | \(1.032 \pm 0.026\) (0.98–1.08) | 57.8 |
DCRNN | \(0.0364 \pm 0.0007\) (0.0350–0.0378) | \(0.857 \pm 0.021\) (0.82–0.90) | 23.6 |
DCRNN6G | \(0.0367 \pm 0.0009\) (0.0349–0.0385) | \(0.869 \pm 0.024\) (0.83–0.91) | 72.4 |
