Dynamic Graph Convolutional Recurrent Network for Traffic Prediction: Benchmark and Solution

Abstract

Traffic prediction is the cornerstone of intelligent transportation system. Accurate traffic forecasting is essential for the applications of smart cities, i.e., intelligent traffic management and urban planning. Although various methods are proposed for spatio-temporal modeling, they ignore the dynamic characteristics of correlations among locations on road network. Meanwhile, most Recurrent Neural Network (RNN) based works are not efficient enough due to their recurrent operations. Additionally, there is a severe lack of fair comparison among different methods on the same datasets. To address the above challenges, in this paper, we propose a novel traffic prediction framework, named Dynamic Graph Convolutional Recurrent Network (DGCRN). In DGCRN, hyper-networks are designed to leverage and extract dynamic characteristics from node attributes, while the parameters of dynamic filters are generated at each time step. We filter the node embeddings and then use them to generate dynamic graph, which is integrated with pre-defined static graph. As far as we know, we are first to employ a generation method to model fine topology of dynamic graph at each time step. Further, to enhance efficiency and performance, we employ a training strategy for DGCRN by restricting the iteration number of decoder during forward and backward propagation. Finally, a reproducible standardized benchmark and a brand new representative traffic dataset are opened for fair comparison and further research. Extensive experiments on three datasets demonstrate that our model outperforms 15 baselines consistently. Source codes are available at https://github.com/tsinghua-fib-lab/Traffic-Benchmark.

Residual Graph Convolutional Recurrent Networks For Multi-step Traffic Flow Forecasting

“…• DGCRN [38]: The model proposes a supernetwork with adaptive stepby-step generation of dynamic adjacency matrices, which significantly improves the forecasting performance.…”

“…GWNET-conv [37] introduces a new covariance loss on Graph WaveNet to significantly improve the forecasting accuracy of Graph WaveNet. DGCRN [38] uses a supernetwork to generate the adjacency matrix and merge it with the original road network matrix to capture spatial correlations dynamically. STGNN [39] model provides a learnable spatial graph neural network of location attention mechanisms and captures local and global temporal correlations using GRU and transformer layers.…”

Pre-training Enhanced Spatial-temporal Graph Neural Network for Multivariate Time Series Forecasting

“…S 𝑖 𝑡𝑟𝑎𝑖𝑛 is static for all samples during training, helping to make the training process more robust and accurate. The feature H 𝑖 is dynamic for different training samples to reflect the dynamics of dependency graphs [19]. As such, we use the cross-entropy between Θ and the 𝑘NN graph 𝐴 𝑎 as graph structure regularization:…”

Decoupling Long- and Short-Term Patterns in Spatiotemporal Inference

“…Inductive Adaptive Graph Generation To model the hidden relations among nodes, (Wu et al 2019;Bai et al 2020;Wu et al 2020) learn a static adaptive adjacency matrix but disregard the dynamic dependencies. Later, (Li et al 2021) handle the dynamic relations by learning the matrix at each recurrent step. However, these methods significantly depend on the node embedding layer which is not available in the inductive settings.…”

“…For instance, (Wu et al 2020) propose a dilated inception temporal convolution to discover relations with different scales and enlarge receptive fields of the model for long sequence modeling. To address hidden relations among sensors, (Li et al 2021;Bai et al 2020) further propose graph generation modules that learn an adaptive adjacency matrix to describe the hidden relation strength between two nodes. However, these models are based on transductive learning, which cannot handle a variable number of input nodes and are less explored for inference problems.…”