GMAN: A Graph Multi-Attention Network for Traffic Prediction
By Chuanpan Zheng, Xiaoliang Fan, Cheng Wang, Jianzhong Qi. AAAI Conference on Artificial Intelligence, 2019
1,978 citations recorded by the source corpus.
Long-term traffic prediction is highly challenging due to the complexity of traffic systems and the constantly changing nature of many impacting factors. In this paper, we focus on the spatio-temporal factors, and propose a graph multi-attention network (GMAN) to predict traffic conditions for time steps ahead at different locations on a road network graph. GMAN adapts an encoder-decoder architecture, where both the encoder and the decoder consist of multiple spatio-temporal attention blocks to model the impact of the spatio-temporal factors on traffic conditions. The encoder encodes the input traffic features and the decoder predicts the output sequence. Between the encoder and the decoder, a transform attention layer is applied to convert the encoded traffic features to generate the sequence representations of future time steps as the input of the decoder. The transform attention mechanism models the direct relationships between historical and future time steps that helps to alleviate the error propagation problem among prediction time steps. Experimental results on two real-world traffic prediction tasks (i.e., traffic volume prediction and traffic speed prediction) demonstrate the superiority of GMAN. In particular, in the 1 hour ahead prediction, GMAN outperforms state-of-the-art methods by up to 4% improvement in MAE measure. The source code is available at https://github.com/zhengchuanpan/GMAN.
Subfields: Building and Construction, Control and Systems Engineering
View Paper: https://www.semanticscholar.org/paper/fc41d75288a81dd7e30087480f51821fc6572d95
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Building and Construction
- Junbo Zhang
- Jiayun Luo
- Weiwei Jiang
- Yu Zheng
- Nizar Bouguila
- Saeed Rahmani
- Asiye Baghbani
- Zachary Patterson
In Control and Systems Engineering
- Alvaro Sanchez-Gonzalez
- Todd Hester
- Weiwei Jiang
- Geoffrey I. Webb
- Nizar Bouguila
- Cesare Alippi
- Peter W. Battaglia
- Bryan Hooi
Citations in the corpus
- Decoupled Dynamic Spatial-Temporal Graph Neural Network for Traffic Forecasting (2022, neutral): “, traffic flow1) based on historical traffic conditions [51] observed by sensors [9, 19].”
- Spatial-Temporal Identity: A Simple yet Effective Baseline for Multivariate Time Series Forecasting (2022, methodology): “We select a wealth of baselines that have official public code, including the traditional methods (VAR [12], HI [4]) and the typical deep learningmethods (LSTM [6], DCRNN [10], STGCN [20], Graph WaveNet [18], AGCRN [1], StemGNN [2]), as well as the very recent works (GMAN [22], MTGNN [17], ST-Norm [”
- Spatio-Temporal Self-Supervised Learning for Traffic Flow Prediction (2022, background): “…flow prediction with various deep learning techniques, e.g., convolutional neural networks over region grids (Zhang, Zheng, and Qi 2017), graph neural networks for spatial dependency modeling (Zhang et al. 2021), and attention mechanism for spatial information aggregation (Zheng et al. 2020).”
- Pre-training Enhanced Spatial-temporal Graph Neural Network for Multivariate Time Series Forecasting (2022, methodology): “In fact, most models use historical data in a small window to make predictions, e.g., use the past twelve time steps (one hour) to predict the future twelve time steps [20, 29, 35, 36, 41].”
- Graph Neural Networks for Intelligent Transportation Systems: A Survey (2023, background): “To address the multi-step (and relatively long-term) prediction of traffic conditions, Zheng et al. [36] proposed GMAN, a graph multi-attention network, which adopts a spatial-temporal encoder-decoder architecture coupled with a transform attention layer between the encoder and decoder.”
- Deep Learning on Traffic Prediction: Methods, Analysis, and Future Directions (2021, methodology): “…current traffic condition and the previous observations at a certain position in the road network, but also model the long-term sequence data to solve the deficiencies of RNN. [62] designed a temporal attention mechanism to adaptively model the non-linear correlations between different time steps.”
- Spatio-Temporal Graph Neural Networks for Predictive Learning in Urban Computing: A Survey (2023, methodology): “Model Specific Task Spatial learning Temporal Learning Fusion Architecture STGCN [18] Traffic state prediction Spectral GCN Gated TCN Factorized T-GCN [24] Traffic state prediction Spatial GCN GRU Factorized DCRNN [26] Traffic state prediction Spatial GCN GRU Coupled GMAN [118] Traffic state predict”
- A Survey on Graph Neural Networks for Time Series: Forecasting, Classification, Imputation, and Anomaly Detection (2023, background): “For instance, GMAN [70] attentively aggregates historical information by considering both spatial and temporal features.”
- Graph Neural Network for Traffic Forecasting: A Survey (2021, background): “Problem Relevant Studies Road Traffic Flow [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62],”
- PDFormer: Propagation Delay-aware Dynamic Long-range Transformer for Traffic Flow Prediction (2023, background): “GMAN and ASTGNN retain a time-consuming encoder-decoder structure, which is replaced by a forward procedure in STTN and PDFormer.”
- Spatio-Temporal Adaptive Embedding Makes Vanilla Transformer SOTA for Traffic Forecasting (2023, methodology): “Recent models, including PDFormer [13], GMAN [41] and STID [24], apply spatial embedding 𝑠 .”
- Graph Neural Network-Based Anomaly Detection in Multivariate Time Series (2021, methodology): “Related variants have shown success in time-dependent problems: for example, GNN-based models can perform well in traffic prediction tasks (Yu, Yin, and Zhu 2017; Chen et al. 2019; Zheng et al. 2020).”
- An Attentive Survey of Attention Models (2019, background): “Zhu et al. [2021] further use deformable attention module to reduce the training time of DETR by 10×, lowering its computational cost.”
- Dynamic and Multi-faceted Spatio-temporal Deep Learning for Traffic Speed Forecasting (2021, methodology): “Most of the recent research has focused on geometric spatio-temporal learning [1, 14, 19, 26, 27, 31, 32, 35].”
- Dynamic Graph Convolutional Recurrent Network for Traffic Prediction: Benchmark and Solution (2021, methodology): “CNN [17, 32, 33, 35], GNN [1, 4, 6–9, 12, 15, 18, 23, 28–30, 34, 36, 37], and attention mechanism [4, 8, 9, 12, 16, 18, 19, 31, 38] have been widely used to capture spatial dependency.”
- ETA Prediction with Graph Neural Networks in Google Maps (2021, methodology): “This generic blueprint of a spatial GNN component combined with a temporal component yielded several subsequent proposals, including STGCN [34], GaAN [37], Graph WaveNet [29], ST-GRAT [19], StemGNN [3], and GMAN [39].”