Graph Neural Rough Differential Equations for Traffic Forecasting

20 Mar 2023  ·  Jeongwhan Choi, 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 rough differential equation (STG-NRDE). Neural rough differential equations (NRDEs) are a breakthrough concept for processing time-series data. Their main concept is to use the log-signature transform to convert a time-series sample into a relatively shorter series of feature vectors. We extend the concept and design two NRDEs: 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 27 baselines. STG-NRDE shows the best accuracy in all cases, outperforming all those 27 baselines by non-trivial margins.

PDF Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Traffic Prediction PeMSD3 STG-NRDE 12 steps MAE 15.50 # 5
12 steps RMSE 27.06 # 4
12 steps MAPE 14.9 # 4
Traffic Prediction PeMSD4 STG-NRDE 12 steps MAE 19.13 # 11
12 steps RMSE 30.94 # 5
12 steps MAPE 12.68 # 2
Traffic Prediction PeMSD7 STG-NRDE 12 steps MAE 20.45 # 7
12 steps RMSE 33.73 # 2
12 steps MAPE 8.65 # 2
Traffic Prediction PeMSD7(L) STG-NRDE 12 steps MAE 2.85 # 5
12 steps RMSE 5.76 # 3
12 steps MAPE 7.14 # 2
Traffic Prediction PeMSD7(M) STG-NRDE 12 steps MAE 2.66 # 5
12 steps RMSE 5.31 # 5
12 steps MAPE 6.68 # 3
Traffic Prediction PeMSD8 STG-NRDE 12 steps MAE 15.32 # 11
12 steps RMSE 24.72 # 6
12 steps MAPE 8.9 # 3

Methods