Table 1 Taxonomy of classical and shallow-learning models.
From: 6G conditioned spatiotemporal graph neural networks for real time traffic flow prediction
Model | Spatial handling | Temporal model | Key strengths/limitations |
|---|---|---|---|
None | Linear autoregression | Simple, interpretable; no spatial coupling | |
Kalman Filter14 | None | State-space dynamics | Good for noisy data; spatially agnostic |
SVR17 | Manual features | Kernel regression | Nonlinear; requires feature engineering |
Random Forest18 | Manual features | Ensemble of trees | Robust to noise; no end-to-end spatial learning |
GBM/XGBoost19 | Manual features | Gradient boosting | High accuracy; long training, no explicit graph |
