Jeongwhan Choi, Hwangyong Choi, Jeehyun Hwang, Noseong Park
Traffic forecasting is one of the most popular spatio-temporal tasks in the field of machine learning. A prevalent approach in the field is to combine graph convolutional networks and recurrent neural networks for the spatio-temporal processing. There has been fierce competition and many novel methods have been proposed. In this paper, we present the method of spatio-temporal graph neural controlled differential equation (STG-NCDE). Neural controlled differential equations (NCDEs) are a breakthrough concept for processing sequential data. We extend the concept and design two NCDEs: one for the temporal processing and the other for the spatial processing. After that, we combine them into a single framework. We conduct experiments with 6 benchmark datasets and 20 baselines. STG-NCDE shows the best accuracy in all cases, outperforming all those 20 baselines by non-trivial margins.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Traffic Prediction | PeMSD7(M) | 12 steps MAE | 2.68 | STG-NCDE |
| Traffic Prediction | PeMSD7(M) | 12 steps MAPE | 6.76 | STG-NCDE |
| Traffic Prediction | PeMSD7(M) | 12 steps RMSE | 5.39 | STG-NCDE |
| Traffic Prediction | PeMSD4 | 12 steps MAE | 19.21 | STG-NCDE |
| Traffic Prediction | PeMSD4 | 12 steps MAPE | 12.76 | STG-NCDE |
| Traffic Prediction | PeMSD4 | 12 steps RMSE | 31.09 | STG-NCDE |
| Traffic Prediction | PeMSD7(L) | 12 steps MAE | 2.87 | STG-NCDE |
| Traffic Prediction | PeMSD7(L) | 12 steps MAPE | 7.31 | STG-NCDE |
| Traffic Prediction | PeMSD7(L) | 12 steps RMSE | 5.76 | STG-NCDE |
| Traffic Prediction | PeMSD8 | 12 steps MAE | 15.45 | STG-NCDE |
| Traffic Prediction | PeMSD8 | 12 steps MAPE | 9.92 | STG-NCDE |
| Traffic Prediction | PeMSD8 | 12 steps RMSE | 24.81 | STG-NCDE |
| Traffic Prediction | PeMSD7 | 12 steps MAE | 20.53 | STG-NCDE |
| Traffic Prediction | PeMSD7 | 12 steps MAPE | 8.8 | STG-NCDE |
| Traffic Prediction | PeMSD7 | 12 steps RMSE | 33.84 | STG-NCDE |
| Traffic Prediction | PeMSD3 | 12 steps MAE | 15.57 | STG-NCDE |
| Traffic Prediction | PeMSD3 | 12 steps MAPE | 15.06 | STG-NCDE |
| Traffic Prediction | PeMSD3 | 12 steps RMSE | 27.09 | STG-NCDE |