Adaptive Graph Convolutional Recurrent Network for Traffic Forecasting

NeurIPS 2020  ·  Lei Bai, Lina Yao, Can Li, Xianzhi Wang, Can Wang ·

Modeling complex spatial and temporal correlations in the correlated time series data is indispensable for understanding the traffic dynamics and predicting the future status of an evolving traffic system. Recent works focus on designing complicated graph neural network architectures to capture shared patterns with the help of pre-defined graphs. In this paper, we argue that learning node-specific patterns is essential for traffic forecasting while the pre-defined graph is avoidable. To this end, we propose two adaptive modules for enhancing Graph Convolutional Network (GCN) with new capabilities: 1) a Node Adaptive Parameter Learning (NAPL) module to capture node-specific patterns; 2) a Data Adaptive Graph Generation (DAGG) module to infer the inter-dependencies among different traffic series automatically. We further propose an Adaptive Graph Convolutional Recurrent Network (AGCRN) to capture fine-grained spatial and temporal correlations in traffic series automatically based on the two modules and recurrent networks. Our experiments on two real-world traffic datasets show AGCRN outperforms state-of-the-art by a significant margin without pre-defined graphs about spatial connections.

PDF Abstract NeurIPS 2020 PDF NeurIPS 2020 Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Uses Extra
Training Data
Result Benchmark
Traffic Prediction BJTaxi AGCRN MAE @ in 12.30 # 2
MAE @ out 12.38 # 2
MAPE (%) @ in 15.61 # 2
MAPE (%) @ out 15.75 # 2
Weather Forecasting LA AGCRN MSE (t+1) 0.2289 ± 0.0327 # 2
MSE (t+6) 0.8412 ± 1.1162 # 3
Traffic Prediction NE-BJ AGCRN 12 steps MAE 4.99 # 5
Weather Forecasting NOAA Atmospheric Temperature Dataset AGCRN MAE (t+1) 0.3019 ± 0.0374 # 3
MAE (t+10) 1.3755 ± 0.2732 # 2
Traffic Prediction NYCBike1 AGCRN MAE @ in 5.17 # 2
MAE @ out 5.47 # 2
MAPE (%) @ in 25.59 # 2
MAPE (%) @ out 26.63 # 2
Traffic Prediction NYCBike2 AGCRN MAE @ in 5.18 # 2
MAE @ out 4.79 # 2
MAPE (%) @ in 27.14 # 2
MAPE (%) @ out 26.17 # 2
Traffic Prediction NYCTaxi AGCRN MAE @ in 12.13 # 2
MAE @ out 9.87 # 2
MAPE (%) @ in 18.78 # 2
MAPE (%) @ out 18.41 # 2
Traffic Prediction PeMS04 AGCRN 12 Steps MAE 19.83 # 9
Weather Forecasting SD AGCRN MSE (t+1) 0.2010 ± 0.0188 # 2
MSE (t+6) 1.0181 ± 0.1275 # 9

Methods